2724 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011

Size: px
Start display at page:

Download "2724 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011"

Transcription

1 2724 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 SAR Image Despeckling Based on Local Homogeneous-Region Segmentation by Using Pixel-Relativity Measurement Hongxiao Feng, Student Member, IEEE, Biao Hou, Member, IEEE, and Maoguo Gong, Member, IEEE Abstract This paper provides a novel pointwise-adaptive speckle filter based on local homogeneous-region segmentation with pixel-relativity measurement. A ratio distance is proposed to measure the distance between two speckled-image patches. The theoretical proofs indicate that the ratio distance is valid for multiplicative speckle, while the traditional Euclidean distance failed in this case. The probability density function of the ratio distance is deduced to map the distance into a relativity value. This new relativity-measurement method is free of parameter setting and more functional compared with the Gaussian kernelprojection-based ones. The new measurement method is successfully applied to segment a local shape-adaptive homogeneous region for each pixel, and a simplified strategy for the segmentation implementation is given in this paper. After segmentation, the maximum likelihood rule is introduced to estimate the true signal within every homogeneous region. A novel evaluation metric of edge-preservation degree based on ratio of average is also provided for more precise quantitative assessment. The visual and numerical experimental results show that the proposed filter outperforms the existing state-of-the-art despeckling filters. Index Terms Homogeneous-region speckle-product model, local homogeneous-region segmentation, pixel-relativity measurement, synthetic aperture radar (SAR) image despeckling. I. INTRODUCTION SYNTHETIC aperture radar (SAR) images contain abundant terrain information for extensive applications in ground and sea monitoring, archeology, disaster assessment, and so on. Speckle is an inherent phenomenon accompanying SAR image for the coherent imaging property of SAR. Sometimes, speckle is useful, such as in SAR-based glacier-flow monitoring. However, in many cases, for example, image clas- Manuscript received July 28, 2009; revised June 25, 2010 and November 23, 2010; accepted January 9, Date of publication March 10, 2011; date of current version June 24, This work was supported in part by the National Natural Science Foundation of China ( , , , , and ), by the China Postdoctoral Science Foundation funded Project ( ), by the National Research Foundation for the Doctoral Program of Higher Education of China ( ), by the Program for Cheung Kong Scholars and Innovative Research Team in the University of China under Grant IRT0645, by the China Postdoctoral Science Foundation Special funded Project ( ), and by the National High Technology Research and Development Program (863 Program) of China (2009AA12Z210). The authors are with the Key Laboratory of Intelligent Perception and Image Understanding of the Ministry of Education, Xidian University, Xi an , China ( hxfeng@mail.xidian.edu.cn; avcodec@163.com; gong@ieee.org). Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /TGRS sification and object detection, the existence of speckle leads to the degradation of image quality and makes an undesired effect on the SAR images visual and automatic interpretation. Therefore, speckle removal is a key and indispensable step in SAR image preprocessing. In this paper, we treat the speckle as a noise and consider that the SAR image consists of true terrain backscatterer and speckle from the viewpoint of despeckling. The purpose of despeckling is to suppress speckle and restore the true backscatterer. Generally, the most popular speckle-suppression methods could be classified into two groups: multilook processing and filtering methods. Multilook processing reduces the variance of speckle while degrading the image spatial resolution. Compared with multilook processing, filtering methods could avoid this shortage to a certain degree. During the past several decades, plenty of filtering methods have been presented to suppress speckle. Among these filters, many classical filters have been widely applied, including the following: Lee filter [1], Kuan filter [2], Frost filter [3], Gamma-MAP filter [4], enhanced Lee or Frost filter [5], improved Sigma filter [6], and so on. These classical filters assume that the true backscatterer value is a local stationary random process. In fact, the assumption does not completely hold true for the case of a local sliding square window, including strong edge or texture, and thus, this causes oversmoothing in the texture or edge regions. Although the series of enhanced filters could effectively preserve texture and edge, it is difficult to distinguish homogeneous regions from texture or edge accurately. Wavelet has been theoretically proven effective in nonstationary signal analysis and is also applied in SAR image despeckling [7] [10]. The wavelet-based methods can preserve edge and texture but are prone to bring noticeable artifacts, such as Gibbs-like ringing in uniform and edge in nearby areas [11], [12] due to the own drawbacks of the wavelet and the similar frequency characteristics of noise and edge. As a modified version, undecimated wavelet transform holds a translation-invariant property. The despeckling approaches based on it obtain remarkable results [13] [15] and obviously restrain the artifacts in uniform regions. Under certain conditions, the nonstationary issue can be transformed into stationary issue. Within a local homogeneous region, the true backscatter value is a constant [5], and then the local scene can be treated as a stationary scene. As a curve is piecewise approximated by short lines, a SAR image consists of many small homogeneous regions which vary in sizes and shapes, and a single pixel is the minimum size among /$ IEEE

2 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2725 these regions. If the homogeneous regions could be determined successfully, the nonstationarity issue will be transformed into be a stationary one. Accordingly, SAR image despeckling can be treated as a simple linear-image restoration within these regions. This framework can be summarized in two steps: local homogeneous-region segmentation and estimation of the true signal with the segmented results. Homogeneous-region segmentation is a key issue for SAR image despeckling. Lopes et al. [5] addressed that SAR image scene is segmented based on the heterogeneity of scene and suggested that the scene is divided into three classes: homogeneous, heterogeneous, and extremely heterogeneous scene, by using variation coefficient. This approach has been widely applied in despeckling methods, such as the enhanced Lee and Frost filters [5]. Applying the assumption of gammadistribution to homogeneous scenes, a similar method of scene segmentation is operated using a chi-square test instead of variation coefficient [16]. Argenti et al. proposed a scene-segmentation method [17] in the wavelet domain with the clean wavelet coefficients. An improved segmentation method of [17] is presented in [14], which is implemented in undecimated wavelet domain and segments the coefficients into homogeneous and different-level heterogeneous classes with their texture energy. These aforementioned approaches consider the scene segmentation globally without exception and only make a coarse segmentation for homogeneous regions. A refined method of homogeneous-region segmentation is proposed in [18], which determines a homogeneous region for every pixel by using region variance. The local equivalent number of looks (ENL) estimation [19] is applied in the improved version of [18]. The methods in [18] and [19] segment image into many overlapping local homogeneous regions for every pixel. They are consistent with the ideas of classical filters, such as the Lee filter, and apply a sliding square window with varying size based on local-scene properties, instead of a fixed one. Through the use of varying sliding window, the local stationary assumption for classical speckle filters can be satisfied better within the window. However, the homogeneous regions segmented by these methods are usually rectangular and not accurate, particularly for texture or edge areas, which make these filters only effective on expanded homogeneous regions, such as expanded agricultural areas. It is important to describe the relativity between two noisy pixels robustly in image denoising, segmentation, or classification. Buades et al. [20] presented an approach of relativity measurement between two pixels utilizing the Euclidean distance between two noisy image patches and designed non local mean (NL-means in [20]) filter based on this effective and robust measurement method. Unfortunately, the method proposed in [20] is effective only in handling additive noise but is invalid in the case of multiplicative noise [21]. In order to solve this problem, Coupe et al. [22] proposed a kind of relativity-measurement method for speckled image. It is assumed that the noise is a mixture of additive and multiplicative noise, and the multiplicative noise obeys Gaussian distribution. Based on these assumptions, the Pearson distance is defined to measure the relativity of two image patches. Another distance is defined for the multiplicative model in [21], which describes the relativity through a ratio of Euclidean distance to image patch product. Deledalle et al. [23] developed an iterative approach of relativity measurement, and the relativity is designed with a combination of two distances: the distance between current restored-image patches and the distance between noisy patches which is similar to the method in [21]. By utilizing prior knowledge, this method improves the precision of relativity measurement compared with [21] and [22] and is shown to be an absorbing relativity-measurement framework. This paper proposed a segmentation-based despeckling approach for SAR images. The authors focus on segmenting a local homogeneous region for every pixel, which is inspired by the works in [18] and [19]. Ratio distance is developed to represent the distance between two speckled-image patches. The work in [21] gives us the idea to research the probability density function (PDF) of ratio distance. By utilizing this PDF, the ratio distance is mapped into a relativity value which is utilized to segment a local homogeneous region for the interested pixel. Maximum likelihood (ML) rule is applied to estimate the true signal within the obtained homogeneous region. The whole approach consists of local region segmentation and true-signal estimation, which is well consistent with the aforementioned two-step framework. Edge-preservation ability plays an important role in the performance evaluation of despeckling filters. Edge-preservation index (EPI) was introduced in [24] and has been applied in [25] [27], which calculates the spatial edges by absolute differences of two adjacent pixels along some direction. It is demonstrated in this paper that EPI is not an effective tool to evaluate the edge preservation for real SAR image and can only be employed for simulated speckle image (i.e., the real image should beforehand be obtained). Inspired by the idea of the ratio distance, a new metric for the evaluation of edge preservation is presented, and its rationality is shown in theory and experiments. This paper is mainly composed of three parts: pixel-relativity measurement, local homogeneous-regions segmentation, and ML estimation. The remainder of this paper is organized as follows. In Section II, the basic speckle statistical characteristics and multiplicative model in homogeneous region are introduced for multilook amplitude SAR image. A novel definition of pixel relativity for SAR image under multiplicative model is developed in Section III. The segmentation of homogeneous region and the ML estimation for true signal are exhibited in Sections IV and V. Section VI explains the proposed edgepreservation evaluation metric and provides the experimental results for real and simulated SAR images, and the conclusion is made in Section VII. II. SPECKLE STATISTICAL CHARACTERISTICS Under the assumption of fully developed speckle [28], an L-look amplitude SAR image obeys the generalized gamma distribution (GGD) [29, p. 108] (also named Nakagami Rayleigh distribution [23]) p A (A) =2 ( ) L ) L 1 ( R Γ(L) exp LA2 A (2L 1), A 0. R (1)

3 2726 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 Here, L represents the number of looks, A indicates the observed SAR image. R represents the underlying radar reflectivity, i.e., true backscattering terrain factor. The m-order moment of GGD is equal to R m/2 Γ(L + m/2)/(l m/2 Γ(L)). Inthis paper, we will only discuss the amplitude image. Multilook amplitude SAR speckle is multiplicative [31], which can be described as Y = X Z (2) where Y represents the observed signal, X is the underlying reflectivity for the amplitude SAR image and is equal to R, and Z indicates an uncorrelated multiplicative speckle. X depends on the properties of the ground and is considered to be a nonstationary random process, while Z is stationary. Z can be treaded as white noise [3], [29, p. 153] with a unit mean. Generally, an image could be divided into many uniform regions that vary in shapes and sizes. In this paper, it is assumed that the SAR image consists of many homogeneous regions, and a single pixel is the minimum case of these regions. Since X is a constant within a homogeneous region [5], the product model is rewritten as Y Ω = C Ω Z Ω (3) where Ω represents a homogeneous region and C Ω represents a constant that is dependent on Ω. It is worth noticing that two interesting conclusions can be obtained from the simplified product model E[Y Ω ]=C Ω E[Z Ω ]=C Ω E[Z] =C Ω = R Ω σ 2 Z Ω = σ2 Y Ω C 2 Ω = σ2 Y Ω E 2 [Y Ω ] = LΓ2 (L) Γ 2 1. (4) (L +0.5) These conclusions show that, within a homogeneous region, the randomness of the observed signal is only caused by a speckle, and the speckle variance is an invariant value that is dependent on the number of looks. III. PIXEL-RELATIVITY MEASUREMENT Image is 2-D signal, and the pixel value represents the digital observed value of this 2-D signal in spatial domain. For the whole image, these pixels can be marked with different classes, i.e., image segmentation, while the relativity measurement between these pixels plays an essential role in image segmentation. The determination of homogeneous regions could be treated as a special issue of image segmentation. It goes without saying that relativity measurement is crucial to homogeneousregion segmentation. A. Traditional Measurement of Pixel Relativity and Problem Plenty of distance measurements are developed to represent the relativity of two samples in the field of pattern classification. Between these distances, Euclidean distance is an essential and widely used measurement in pattern recognition and classification. In practice, Euclidean distance is usually mapped by a Gaussian kernel radial basis function ( ) R i =exp β P 0 P i 2 2 (5) where P 0 is the central pixel value and P i represents the pixel value on location i. R i represents the relativity between the two pixels. The parameter β acts as a degree of kernel function, which controls the decay of the function. The definition of (5) is used in many neighborhood-averaging filters, for example, the Similar Univalue Segment Assimilating Nucleus (SUSAN) filter [32] and the Bilateral filter [33]. Definition (5) can supply a reliable relativity estimation of two pixels for noise-free image. However, SAR images are contaminated by speckle, and thus, (5) becomes unreliable. A possible solution is to expand the sample capacity. In practice, the Euclidean distance of P 0 and P i is usually replaced by the Euclidean distance between two square-image patches N 0 and N i, which contain P 0 and P i in their centers, respectively, i.e., the distance is developed for NL-means filter [20] (we only consider the case of SAR image here) ( ) R i =exp β Y N0 Y Ni 2 2,G (6) where Y N0 and Y Ni denote two patches which have the same size and shape in the observed SAR image Y, and G represents the standard Gaussian kernel function. Unfortunately, (6) is only robust for additive noise and no more effective for multiplicative noise, which was mentioned in [21]. However, a theoretical proof of the invalidity is absent from the thesis. Here, we will make a detailed proof as follows. Considering convenience of demonstration, the product model can be transformed into an additive model with a signaldependent noise [9], [13], [14] Y = X Z = X + X(Z 1) = X + S. (7) Let M ΔY = Y N0 Y Ni 2 2 = Y N0 (k) Y Ni (k) 2 where M denotes the number of pixels of image patches Y N0 and Y Ni. The expectation is calculated as E[ΔY ]= = = = E ( Y N0 (k) Y Ni (k) 2) ( ( E XN0 (k) X Ni (k) ) + ( S N0 (k) S Ni (k) ) 2) E ( ΔX(k)+ΔS(k) 2) { [ΔX(k)] 2 +2ΔX(k)E [ΔS(k)] + E ΔS(k) 2} E[ΔS(k)]=0 = ΔX + E ΔS(k) 2. (8)

4 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2727 Here E ΔS(k) 2 = E S N0 (k) S Ni (k) 2 = [ E S N0 (k) 2 2E [S N0 (k) S Ni (k)] + E S Ni (k) 2]. Considering the white-noise assumption for speckle, Z N0 (k) is independent of Z Ni (k)(i 0) E [S N0 (k) S Ni (k)] and thus = E [X N0 (k)(z N0 (k) 1) X Ni (k)(z Ni (k) 1)] = X N0 (k) X Ni (k)e [Z N0 (k) 1] E [Z Ni (k) 1] = 0 E ΔS(k) 2 = [E S N0 (k) 2 + E S Ni (k) 2] Thus we get = =σz 2 { [X N0 (k)] 2 +[X Ni (k)] 2} E(Z 1) 2 E[ΔY ]= X N0 X Ni σ2 Z { [X N0 (k)] 2 +[X Ni (k)] 2}. ( ) X N X N i 2 2. (9) Equation (9) shows that the Euclidean distance is robust when the true signal X is a constant, i.e., it is only robust for ideal uniform regions of SAR image. However, the second term of (9) will vary according to local scenes in edge and texture regions, and therefore, (6) will no longer be valid. B. Ratio Distance for Multiplicative Noise Model The distinction between two image patches could not only be denoted as the difference of these patches but could also be denoted as the ratio of the patches. In this paper, a universal ratio distance is developed the idea of which arises from the ratio of average (ROA) operator [34]. Let symbol./ denote the dot quotient of two vectors; the ratio distance is then defined as M d i = Y Ni./Y N0 2 2,G = G(k) Y Ni (k) 2 Y N0 (k), Y N0 (k) 0. (10) Although the aforementioned definition is simple, it is very robust for multiplicative noise. The following formula will show the validity proof of the ratio distance in theory. Due to the assumption of white noise for speckle, we obtain E[ Z Ni (k)/z No (k) 2 ]=E Z Ni (k) 2 /E Z No (k) 2, and so E[d i ]= G(k)E Y Ni (k) 2 Y N0 (k) = G(k) X Ni (k) 2 E Z Ni (k) 2 X N0 (k) E Z No (k) 2 = G(k) X Ni (k) 2 ( σz 2 +1) X N0 (k) (σ 2 Z +1) = X Ni./X N0 2 2,G. (11) Equation (11) shows the validity and robustness of the ratio distance since, as expected, it conserves the true relativity between pixels. C. Relativity Measured by the PDF of Ratio Distance The ratio distance cannot be directly applied as a relativity after a mapping of Gaussian kernel function because of its property of relativity representation: A ratio distance that is closer to one involves a bigger relativity, while farther from one involves a smaller relativity. Therefore, the ratio distance should be mapped as a rational normalized relativity. Azzabou mapped the Euclidean distance into relativity with the PDF of the distance [21], which gives us the enlightenment to resolve the previous problem. Here, we set r i,k = Y N i (k) Y N0 (k). (12) Expression (12) contains true signal ratio, which make it difficult to handle. Here, it is assumed that Y Ni (k) and Y N0 (k) are two different observed values of real signal X N0 (k), i.e., Y Ni (k) and Y N0 (k) are located within the same homogeneous region. Thus, (12) can be transformed into r i,k = Z N i (k) Z N0 (k). (13) Let p(r i,k ) represent the PDF of r i,k and set X =1;after that, (2) is substituted into (1). We get the PDF of speckle Z as p Z (Z) = 2LL Γ(L) exp( LZ2 )Z (2L 1), Z 0. (14) For convenience of representation, we set t = Z Ni (k) and s = Z N0 (k). Utilizing (14), we get p(r i,k )= = s p T (sr i,k ) p S (s)ds s 2LL (sr i,k ) 2L 1 Γ(L) 2LL (s) 2L 1 Γ(L) exp ( L(sr i,k ) 2) exp( Ls 2 )ds

5 2728 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 Fig. 1. Variation of p(r i,k ) following the number of looks. ratio distance. The multiplicative noise is assumed as Gaussian distribution in [22], while it is absent in this paper. From the aspects of distance definition, the distance for noisy-image patches in [23] is similar to the definition in [21]; the major difference of [23] and [21] is that the prior knowledge of the middle estimated result is introduced to the final relativity measurement in [23]. Compared with this paper, [23] developed a new kind of framework for relativity calculation, while this paper focuses on the relativity measurement for speckled-image patches. Under certain conditions, our method can be combined into the framework proposed in [23]. The relativity calculation in [23] requires that it manually set the degree parameter of the Gaussian kernel function, while our approach is free of this parameter setting. = 4L2L (r i,k ) 2L 1 } Γ 2 (L) {{ } =α + 0 = α 2L 1 β s 4L 1 exp ( (r i,k ) 2 +1 ) }{{} =β + 0 (2L 1)! = α β 2L 1 = 2(2L 1)! Γ 2 (L) s 2 s 4L 3 exp( βs 2 )ds = + 0 s exp( βs 2 )ds = α ds (2L 1)! 2β 2L (r i,k ) 2L 1. (15) [(r i,k ) 2 2L +1] The study of p(r i,k ) indicates that the maximum value of p(r i,k ) locates at r = (2L 1)/(2L +1), and r approaches one when the number of looks increases gradually. Fig. 1 shows the variation of p(r i,k ) following L. Utilizing the definition of (15), we define the relativity between pixel P 0 and P i as follows: R i = p(r i,k ) 2 2,G = M [ ( )] 2 YNi (k) G(k) p. (16) Y N0 (k) An important progress of the new relativity measurement in (16) is that it is free of parameter control and only depends on the statistical property of the multiplicative noise. Relativity measurement between two speckled images patches has been studied widely [21] [23]. The major contributions of this paper on relativity measurement are the following: 1) A robust and universal ratio distance for two speckled image patches is developed; 2) a new relativity measurement is produced by the PDF of the ratio distance instead of traditional Gaussian kernel projection; and 3) the novel measurement method proposed by us is almost nonparametric, which only requires the setting of the size of the comparison window. The distance proposed in [21] can be seen as a special case of IV. LOCAL HOMOGENEOUS-REGION ESTIMATION WITH PIXEL-RELATIVITY MEASUREMENT As mentioned in Section III, the measurement of pixel relativity for a SAR image provides a tool for image segmentation. In this section, a method of homogeneous-region segmentation will be developed by applying the relativity-measurement approach proposed in the previous section. Let X 0 (n) be the nth true pixel value and Ω H (n) denote a local homogeneous region surrounding X 0 (n). N is the neighborhood of X 0 (n) and is larger than Ω H (n), i.e., Ω H (n) ={X 0 (i) R i >T R,X 0 (i) N} (17) where T R is a relativity threshold and R i is calculated using (16). The neighborhood N is square and is similar to the searching window in the NL-means algorithm [20]. Generally, a complete pixel traverse in neighborhood N could determine a homogeneous region. However, this complete traverse is time consuming. In this paper, we approximate the local homogeneous region defined in (17) with a polygonal region and give a kind of simplified traversal method which determines the polygon. A set of directions is defined as {d k (n) d k (n) =2kπ/D, k =0,...,D 1} for the nth pixel. Starting with the central pixel Y 0 (n) and traversing along a direction d k (n), an optimal scale Sk (n) can be found. The traverse can be seen in Fig. 2. The optimal-scale set {Sk (n),k =0,...,D 1} is determined after traversing along all directions. Based on these optimal scales, a homogeneous region Ω H (n) is determined. Fig. 3 shows an example of a homogeneous region constructed by eight directions. It is worth noticing that the shape of homogeneous region Ω H (n) will be consistent with the case in [18] and [19] when the direction number D is equal to four, and all of them are rectangular. Nevertheless, two novelties of our method are addressed compared with [18] and [19]: 1) The relativity between two noisy pixels is utilized to segment the local homogeneous region, while local statistical characters (local variance or ENL) are applied in [18] and [19], and 2) the homogeneous region constructed by pixel relativity is shape adaptive, and a higher segmentation precision is made compared with a shape-fixed window in [18] and [19].

6 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2729 Fig. 2. Implementing process of determining the optimal-scale set. where K R is a normalized scale factor. T R max represents the maximum value of T R, which can be obtained once the looks of the SAR image is available. We substitute r = (2L 1)/(2L +1) [which corresponds to the maximum value of (15)] into (15) and then get T R max = G(k) 4[(2L 1)!]2 (4L 2 1)(2L 1) 2L 2 Γ 4 (L)(4L) 2L. (19) V. D ESPECKLING WITH ML The multiplicative model is rewritten in homogeneous region Ω H (n) as follows: Y ΩH (n)(k)=c ΩH (n) Z ΩH (n)(k), 1 k M ΩH (n) (20) Fig. 3. One homogeneous region constructed by eight directions. The parameter T R is the most important factor for the segmentation result. Here, it was developed as follows: T R = K R T R max (18) where the total number of pixels within Ω H (n) is M ΩH (n). Considering the GGD of observed signal Y ΩH (n), themlrule is applied to estimate the parameter R of GGD ˆR ΩH (n) =maxl [ Y ΩH (n)(1),y ΩH (n)(2),..., Y ΩH (n) ( MΩH (n)) ; RΩH (n)]. (21)

7 2730 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 The likelihood function is defined as L ( [ R ΩH (n)) = L YΩH (n)(1),y ΩH (n)(2),..., ( ] Y ΩH (n) MΩH (n)) ; RΩH (n) M ΩH (n) ( ) L L 1 = 2 R ΩH (n) Γ(L) exp { L [ Y ΩH (n)(k) ] } 2 R ΩH (n) [ Y ΩH (n)(k) ] (2L 1). (22) By letting f(r ΩH (n)) =lnl(r ΩH (n)), we obtain f ( M ΩH { (n) ) R ΩH (n) = L ln R ΩH(n) L[ Y ΩH (n)(k) ] 2 R ΩH (n) } +ln 2 Γ(L) + L ln L +(2L 1)Y Ω H (n)(k). (23) By setting df (R ΩH (n))/dr ΩH (n)) =0, we get the estimation ˆR ΩH (n) = 1 M ΩH (n) M ΩH (n) [ YΩH (n)(k) ] 2. (24) Equation (24) is an unbiased estimation of ˆR ΩH (n). Since the true reflectivity value is constant within Ω H (n), we get the despeckled result of Y (n) as ˆX(n) = ˆRΩH (n) = 1 M ΩH (n) M ΩH (n) [ YΩH (n)(k) ] 2. (25) VI. EXPERIMENTAL RESULTS AND DISCUSSION In this section, we present the experimental results by applying the proposed approach and compare our method with some classical and recent filters. The performance evaluation of filters is an important and basic issue on SAR image despeckling. As pointed out in [30, pp ], a good despeckling method demonstrates the following properties: 1) speckle reduction in extended uniform regions; 2) edge and texture preservation; 3) absence of artifacts in uniform regions; and 4) radiometric preservation. The performance of despeckling methods can be evaluated based on the aforementioned properties with quantitative indicators and visual impression. A. Novel Evaluation Approach of Edge-Preservation Degree The traditional EPI [24] [27] can be represented as m I D1 (i) I D2 (i) i=1 EPI = m (26) I O1 (i) I O2 (i) i=1 where m is the pixel number of the selected area. I D1 (i) and I D2 (i) represent the adjacent pixel values of the despeckled Fig. 4. Artificial edge images. image along a certain direction. Similarly, I O1 (i) and I O2 (i) represent the adjacent pixel values of the speckled image. As shown in (8) and (9), m i=1 I D1(i) I D2 (i) cannot robustly indicate the edges for the multiplicative-noise model. In this paper, we revised the traditional EPI and present a novel measurement method of edge-preservation degree based on ROA (EPD-ROA) m I D1 (i)/i D2 (i) i=1 EPD-ROA = m. (27) I O1 (i)/i O2 (i) i=1 The validity of EPD-ROA was theoretically revealed in (11). It is similar to EPI; when the EPD-ROA is closer to one, it means better ability of edge preservation. Three artificial edge images (Fig. 4) are contaminated with different-level synthetic speckle, and the edge-preservation degree of these images are calculated with EPI and EPD-ROA. In the experiments, the despeckled images are represented by clean artificial images and by calculating the EPI and the EPD-ROA for the whole image along the horizontal (HD) and vertical (VD) directions. The experimental results are shown in Table I. Since the despeckled image is replaced with clean image, the ideal value of EPI or EPD-ROA should be equal to one. Table I exhibits that EPD-ROA is very robust and close to ideal value of one when the looks is larger than two, while all of the EPI values are far from the ideal value. Although the values of EPD- ROA are not very close to one when L =1or 2, the degree of adjacency is far higher than EPI. It is worth pointing out that EPI and EPD-ROA are all gradually close to the ideal value when the looks increase, but EPD-ROA is stable for different types of edge images and is only affected by the speckle level, whereas EPI varies for all the images. B. Evaluating Index-of-Despeckling Performance In this paper, EPD-ROA and two other different types of quantitative evaluation indicators are applied to evaluate the performance of the different filters. 1) ENL: ENL measures the degree of speckle reduction in a homogeneous region, which is given by [35] ENL = S C ( μhr S C = σ HR ) 2 { 1, intensity image 4/π 1, amplitude image (28) where μ HR and σ HR are the mean and the standard deviation of a chosen homogeneous region, respectively. In practice, we

8 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2731 TABLE I COMPARISON OF MEASUREMENT OF THE EDGE-PRESERVATION DEGREE should artificially choose the homogeneous region to be as large as possible. A large ENL value corresponds to better speckle suppression. 2) Ratio Image: Since the speckle is multiplicative, the ratio of the original SAR image to the despeckled one is a speckle. Therefore, the statistical characteristics of the ratio image could indicate the performance of the despeckled algorithms. It usually evaluates the filters performance by the following two aspects. 1) Mean of ratio image: It measures the degree of radiometric preservation, the ideal value of which is a unit. The mean is closer to one, which means a better performance of radiometric preservation. 2) Variance of ratio image: Under ideal conditions, it is calculated by the following formula [30, pp ]: σ 2 Z = L eγ 2 (L e ) Γ 2 (L e +0.5) 1 (29) where L e represents the effective number of looks [30, pp ]. In this paper, we estimated the effective looks of a real SAR image with the method in [36]. In fact, the aforementioned ideal conditions correspond to a uniform scene. Generally, the real image contains edges and textures, and thus, the genuine variance of the ratio image should be smaller than this ideal value. However, if the value is bigger than the ideal one, (29) can reveal the fact that the despeckling result makes a deviation from the underlying result. C. Discussion of the Parameters for the Proposed Method Several parameters are needed to be set in the proposed method. Among these parameters, the scale factor K R in (18) is the most important parameter, which decides the value of relativity threshold T R and has a serious influence on the despeckled results. The other parameters are the size of image patch N 0 or N i in (16) (we set the size equal to (2f +1) (2f +1)) and the maximal value h k max of scale S k (n) in Section IV. Several groups of experiments are devised to obtain the effect of these parameters on the results. In order to acquire a better shapeadaptive homogeneous region, we set the direction number D as eight and choose the same h max for every scale S k (n). The peak signal-to-noise ratio (PSNR) is used to evaluate the despeckled results. Fig. 5. Test of despeckled result variation following K R. First, we test the parameter K R of three classical natural noise-free images: Peppers, Lena, and Cameraman are corrupted by different-level synthetic speckle. In the experiments, the speckle level is set as L =2, 4, and 8, and the image patch size f and the maximal value h max are empirically set as three and seven. The results are shown as curves in Fig. 5. It is easy to acquire the information from these curves that the threshold K R affects the results seriously, and all the curves contain a peak value. After observing these peak values, an exciting result is obtained that shows that almost all of the maximal values appear as K R within the interval [0.35, 0.4]. This conclusion shows the nicer practicability and operability of our method. In practice, a smaller K R value results in a smoother despeckled result, and a high degree of smoothness corresponds to high level of speckle. Therefore, we suggest that K R be set as for relatively slight speckle, such as the case of L larger than two and as 0.35 for L =1or 2. Second, we set K R =0.375 and f =3. It considers that Lena contains typical uniform and edge regions, which are used to test the influences of parameter h max on the results. In the experiments, noise-free Lena is corrupted by synthetic speckle with L =2, 4, and 8, and we choose h max =4, 5,...,11. Fig. 6(a) shows the relationship between h max and PSNR. The curves in Fig. 6(a) shows that the despeckling results are slightly affected by the value of h max, and the degree of influence decreases when the speckle level decreases. A relatively stable result can be obtained if h max is larger than five. However, the PSNR will decrease clearly in case h max is over eight, because a bigger h max value produces a smoother result. Consequently, we suggest that h max be set as six or seven in practice.

9 2732 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 Fig. 6. Lena. Influence of (a) h k max and (b) f on the despeckling results for speckled Fig. 8. Experimental results for a two-look synthetic curving-edge image. (a) A two-look speckled image ( ). (b) Gamma-MAP. (c) Frost. (d) Wu Maitre. (e)map-uwd-s. (f) LHRS-PRM. Fig. 7. Experimental results for a two-look synthetic regular-edge image. (a) A two-look speckled image ( ). (b) Gamma-MAP. (c) Frost. (d) Wu Maitre. (e) MAP-UWD-S. (f) LHRS-PRM. Finally, the parameter f is tested. It is similar to the test of h max ; Lena is also used in this test and contaminated with L =2, 4, and 8 synthetic speckle. We set K R =0.375 and h max =7, and choose f =1, 2,...,6. The results are shown as curves in Fig. 6(b). These curves reveal that size f has very slight influence on the despeckling results, and a stable result can be obtained once f is more than two. Here, it is notable that the time complexity will increase seriously when f increases. Considering the stability and time cost, we suggest that f should be set as three in practice. Based on aforementioned experiments, it can be known that f is relatively stable, and the degree of smoothness increases as the value of K R or h max decreases or increases, respectively. In contrast, K R has a greater impact on the degree of smoothness than h max. D. Experimental Results for Synthetic and Real SAR Images In this section, three synthetic and four real SAR images are tested; visual and numerical results are obtained for these images. Two of the synthetic images are synthetic regular-edge image [Fig. 4(a)] and synthetic curving-edges image [Fig. 4(b)], which are all corrupted with a two-look speckle and shown in Figs. 7(a) and 8(a), respectively. The third synthetic image is shown in Fig. 9(a) and contaminated with a four-look speckle. The results of our approach (local homogeneous-region segmentation with pixel-relativity measurement, named LHRS- PRM) for these synthetic speckle images have been com- Fig. 9. Experimental results for a four-look synthetic image. (a) A four-look synthetic image ( ). (b) Gamma-MAP. (c) Frost. (d) Wu Maitre. (e) MAP-UWD-S. (f) LHRS-PRM. pared with those of Frost filter [3], Gamma-MAP filter [4], Wu Maitre filter [19], and MAP filter based on undecimated wavelet decomposition and image segmentation (MAP-UWD-S) [14]. In all experiments, the relativity threshold K R was set as 0.375, and the window sizes f of N 0 and N i were all set to three for the relativity calculation. The parameter h max was set as seven for the local homogeneousregion estimation, and the direction number D has been set as eight for all experiments. The sliding window size was 5 5for the Frost and Gamma-MAP filters. The despeckled results for the different algorithms are shown in Figs Table II gives all the numerical evaluation results for these images, which contains the evaluations of PSNR, ratio image, and EPD-ROA along HD and VD for the whole image.

10 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2733 TABLE II NUMERICAL EVALUATION FOR SYNTHETIC SPECKLE IMAGES Fig. 10. Experimental results for Horsetrack. (a) Original real SAR image ( ). (b) Frost. (c) Wu Maitre. (d) MAP-UWD-S. (e) LHRS-PRM. (f) LHRS-PRM-C. Fig. 11. Experimental results for Field. (a) Original real SAR image ( ). (b) Frost. (c) Wu Maitre. (d) MAP-UWD-S. (e) LHRS-PRM. (f) LHRS-PRM-C. The visual comparison of Figs. 7 and 8 indicates that LHRS- PRM is free of artifacts in uniform regions, and the performance smoothness markedly precedes the Gamma-MAP, Frost, and Wu Maitre filter. A very clean uniform region is obtained with our approach, while speckle residue appears in the results obtained with the other filters except with that of MAP-UWD-S. In terms of edge preservation, LHRS-PRM produces clear and sharp edges, while blur phenomenon appears in the results obtained with the Gamma-MAP, Frost, and Wu Maitre filters. Although MAP-UWD-S possesses nicer ability of smoothing, it produces Gibbs-like ringing and edge blurring, particularly in Fig. 7(e). The results shown in Fig. 9 show that LHRS- PRM possesses comparative preserving performance of texture and small objects compared with the Gamma-MAP, Frost, and Wu Maitre filters. MAP-UWD-S could hold a similar preserving performance, whereas edge blurring is unavoidable. Therefore, the finer performance of edge preservation is one of the virtues of our method compared with the other filters. After the presentation of the results for synthetic speckle images, four real SAR images are tested: 1) Ku-band 1-m

11 2734 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 Fig. 12. Experimental results for Yellow River. (a) Original real SAR image ( ). (b) Frost. (c) Wu Maitre. (d) MAP-UWD-S. (e) LHRS-PRM. (f) LHRS-PRM-C. resolution Horse Track near Albuquerque, New Mexico (Fig. 10(a) and named Horsetrack the effective number of looks is about 4); 2) British Defense Research Agency (DRA) SAR X-band 3-m resolution, a rural scene in Bedfordshire, England (Fig. 11(a) and named Field the effective number of looks is about 3.2); 3) Radarsat-2 C-band 8-m resolution, near Yellow River estuary, China (Fig. 12(a) and named Yellow River the effective number of looks is 4); and 4) X-band 3-m resolution in a town near Xi an, China, produced by a Chinese institute (Fig. 13(a) and named Town the effective number of looks is about 4). The comparisons were performed with Frost filter [3], Wu Maitre filter [19], MAP-UWD-S [14], and LHRS-PRM. In the experiments, the setting for all the parameters is completely identical to the setting for the earlier experiments of synthetic images. In Section IV, the authors give a simplified traversal method when one determines a homogeneous region, i.e., the aforementioned approach: LHRS-PRM. Here, we will also show the results obtained by applying complete traversal in the whole spatial neighborhood, named LHRS-PRM-C. In the experiments, its parameter setting is entirely consistent with LHRS-PRM. The readers can download a C-code implementation of LHRS-PRM and LHRS-PRM-C from the website: Figs show the visual results for the four real SAR images. The contrasts for these results indicate that the speckle suppression ability of LHRS-PRM is excellent. The Frost filter produces distinct speckle residue within uniform regions; Fig. 13. Experimental results for Town. (a) Original real SAR image ( ). (b) Frost. (c) Wu Maitre. (d) MAP-UWD-S. (e) LHRS-PRM. (f) LHRS-PRM-C. pointillist-like small regions appear in the results obtained with the Wu Maitre filter. MAP-UWD-S is always accompanied with ringing effect despite its excellent ability of speckle reduction, and the phenomenon is more evident for the Field [Fig. 11(d)]. In contrast, LHRS-PRM effectively removes the speckle in uniform regions, for example, the left area of Horsetrack [Fig. 10(e)] and the middle area of Field [Fig. 11(e)]. Compared with the other filters, LHRS-PRM possesses the best smoothing ability in uniform regions, which is validated by the ENL comparisons in Table III [note: four uniform regions are chosen in Horsetrack and Field, respectively, which are marked in Figs. 10(a) and 11(a)]. In terms of edge and texture preservation, Frost filter can retain the edge and texture features to a certain extent, but the results for uniform regions are not satisfactory. MAP-UWD-S keeps the texture features while resulting in edge blurring. The Wu Maitre filter obtains a tradeoff between smoothing and edge preservation to a certain degree. In contrast, our algorithm preserves the texture well and produces clear edges and without artifacts. Table IV reveals that LHRS-PRM obtained the optimal EPD- ROA in most cases compared with the other algorithms. The radiometric-preservation ability of our approach is comparable with that of the Frost or Wu Maitre filter, but MAP-UWD-S usually produces a relatively larger bias. The visual comparisons indicate that the performance of LHRS-PRM is very close to its completely traverse version: LHRS-PRM-C. It is worth pointing out that LHRS-PRM actually produces sharper edges than LHRS-PRM-C does, which

12 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2735 TABLE III ENL COMPARISON TABLE IV NUMERICAL EVALUATION FOR REAL SAR IMAGES TABLE V IMPLEMENTING-TIME COMPARISON also can be proved through the EPD-ROA comparisons in Table IV. Table V exhibits the contrast of implementing time. It shows that the simplified traverse version possesses noticeably low time cost and is close to one-fifth of the cost of the completely traverse one. Local homogeneous-region segmentation is a critical step for the whole algorithm. A similar idea is also applied in the Wu Maitre filter. In order to compare the segmentation performance between the Wu Maitre filter and LHRS-PRM, a comparative experiment was devised. Because the shape of homogeneous region Ω H (n) is rectangular when the direction number D is equal to four, which is consistent with the case in the Wu Maitre filter, we designed a new method: the segmentation result of the Wu Maitre filter is companied with the ML estimation in this paper. Fig. 14 shows the results for the Wu Maitre-ML filter and the four-directional LHRS- PRM (note: our method can be seen as four-directional LHRS- PRM+ML). It is obvious that an entirely blurring result is produced by the Wu Maitre ML filter, while a very satisfactory result is obtained with the four-directional LHRS-PRM. This fact reveals that the segmentation performance of LHRS-PRM is far higher than the Wu Maitre filter. A drawback of the proposed approach is that it is not very good at processing single-look image since the mapping function is not precise for very high level speckle. This would be solved by making an effective modification on the mapping function when one deals with a single-look image. For computational purpose of the proposed algorithm, it assumes that the maximum size of a homogeneous region is (2S max +1) (2S max +1), the size of window N 0 is M M, the total pixel number of image is N, and the average complexity of the algorithm is [4(S max +1)M 2 +(1+ (2S max +1) 2 )/2]N. Here, 4(S max +1)M 2 N represents the complexity of the homogeneous-region determination, while [1 + (2S max +1) 2 ]N/2 represents the homogeneous-region estimation with ML rule. Since S max and M are commonly

13 2736 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011 PASCO China Corporation for providing the Radarsat-2 data, and the anonymous reviewers and editors for their valuable comments and helpful suggestions which greatly improved this paper s quality. Fig. 14. Visual comparison of the results obtained with (a) Wu Maitre+ML and (b) four-directional LHRS-PRM. far less than N, the complexity level of our algorithm is only O(N). VII. CONCLUSION A novel despeckling methodology has been proposed for SAR image based on local homogeneous-region segmentation with pixel-relativity measurement and ML estimation. The simplified speckle model in homogeneous regions enables the unmanageable multiplicative model to become a simple linear model. The new method of relativity measurement between two pixels only considered the distribution of speckle. It is very robust for speckled SAR image and is almost free of parameter setting and very functional in practice. A homogeneous region is segmented for each pixel by utilizing the proposed pixelrelativity measurement. In particular, the segmentation is accurate for uniform and textured or edged areas. The satisfied estimated result of a true signal is obtained by the ML rule based on the segmented result. Considering the whole algorithm, our method consists of two steps: homogeneous-region segmentation and ML estimation. Classical and recent speckle filters are compared with the proposed approach, and visual impression shows that very clean uniform regions and clear sharp edges are obtained by the proposed method. The numerical results indicate that our method possesses remarkable speckle-suppression ability and good ability of edge and radiometric preservation. It is noticeable that the proposed approach is pointwise; thus, it could easily do parallel processing for software or hardware implementation. ACKNOWLEDGMENT The authors would like to thank T. Bianchi for the warmhearted help of processing the data with the MAP-UDW-S filter, REFERENCES [1] J. S. Lee, Digital image enhancement and noise filtering by use of local statistics, IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-2, no. 2, pp , Mar [2] D. T. Kuan, A. A. Sawchuk, T. C. Strand, and P. Chavel, Adaptive noise smoothing filter for images with signal-dependent noise, IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-7, no. 2, pp , Mar [3] V. S. Frost and J. A. Stiles, A model for radar images and its application to adaptive digital filtering of multiplicative noise, IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-4, no. 2, pp , Mar [4] A. Lopes, E. Nezry, R. Touzi, and H. Laur, Maximum a posteriori filtering and first order texture models in SAR images, in Proc. IGARSS, Washington, DC, 1990, pp [5] A. Lopes, R. Touzi, and E. Nezry, Adaptive speckle filters and scene heterogeneity, IEEE Trans. Geosci. Remote Sens.,vol.28,no.6,pp , Nov [6] J. S. Lee, J. H. Wen, T. L. Ainsworth, K. S. Chen, and A. J. Chen, Improved sigma filter for speckle filtering of SAR imagery, IEEE Trans. Geosci. Remote Sens., vol. 47, no. 1, pp , Jan [7] H. Guo, J. E. Odegard, M. Lang, R. A. Gopinath, I. W. Selesnick, and C. S. Burrus, Wavelet based speckle reduction with application to SAR based ATD/R, in Proc. ICIP, Austin, TX, 1994, vol. 1, pp [8] H. Xie, L. E. Pierce, and F. T. Ulaby, Despeckling SAR images using a low-complexity wavelet denoising process, in Proc. IGARSS, Toronto, ON, Canada, 2002, vol. 1, no. 10, pp [9] D. Gleich and M. Datcu, Gauss Markov model for wavelet-based SAR image despeckling, IEEE Signal Process. Lett., vol. 13, no. 6, pp , Jun [10] D. Gleich and M. Datcu, Wavelet-based despeckling of SAR images using Gauss Markov random fields, IEEE Trans. Geosci. Remote Sens., vol. 45, no. 12, pp , Dec [11] T. Nabil, SAR image filtering in wavelet domain by subband depended shrink? Int. J. Open Probl. Comput. Math., vol. 2, no. 1, pp , Mar [12] Y. Li, J. Yang, L. Sun, and Y. Zhang, SAR speckle reduction based on undecimated tree-structured wavelet transform, in Proc. ICNC, vol. 4222, LNCS, 2006, pp , Part II. [13] F. Argenti and L. Alparone, Speckle removal from SAR images in the undecimated wavelet domain, IEEE Trans. Geosci. Remote Sens., vol. 40, no. 11, pp , Nov [14] T. Bianchi, F. Argenti, and L. Alparone, Segmentation-based MAP despeckling of SAR images in the undecimated wavelet domain, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 9, pp , Sep [15] S. Solbø and T. Eltoft, A stationary wavelet-domain Wiener filter for correlated speckle, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp , Apr [16] C. Wang and R. Wang, Multi-model SAR image despeckling, Electron. Lett., vol. 38, no. 23, pp , Nov [17] F. Argenti, T. Bianchi, and L. Alparone, Multiresolution MAP despeckling of SAR images based on locally adaptive generalized Gaussian pdf modeling, IEEE Trans. Geosci. Remote Sens., vol. 15, no. 11, pp , Nov [18] Y. Wu and H. Maître, Smoothing speckled SAR images by using maximum homogeneous region filters, Opt. Eng., vol. 31, no. 8, pp , Aug [19] J. M. Nicoals, F. Tupin, and H. Maître, Smoothing speckled SAR images by using maximum homogenous filters: An improved approach, in Proc. IGARSS, Sydney, Australia, 2001, vol. 3, pp [20] A. Buades, B. Coll, and J. M. Morel, A non-local algorithm for image denoising, in Proc. CVPR, San Diego, CA, 2005, vol. 2, pp [21] N. Azzabou, Variable bandwidth image models for texture-preserving enhancement of natural images, Ph.D. dissertation, MAS Res. Group (Ecole Centrale de Paris) and DxOLabs, Paris, France, 2008, pp [22] P. Coupé, P. Hellier, C. Kervrann, and C. Barillot, Bayesian non-local means-based speckle filtering, in Proc. IEEE Int. Symp. Biomed. Imaging: From Nano to Macro, Paris, France, May 2008, pp [23] C. Deledalle, L. Denis, and F. Tupin, Iterative weighted maximum likelihood denoising with probabilistic patch-based weights, IEEE Trans. Image Process., vol. 18, no. 12, pp , Dec

14 FENG et al.: SAR IMAGE DESPECKLING BASED ON LOCAL HOMOGENEOUS-REGION SEGMENTATION 2737 [24] J. S. Lee, I. Jurkevich, P. Dewaele, P. Wambacq, and A. Oosterlinck, Speckle filtering of synthetic aperture radar images: A review, Remote Sens. Rev., vol. 8, no. 4, pp , Jan [25] L. Tang, P. Jiang, C. Dai, and J. J. Thomas, Evaluation of smoothing filters suppressing speckle noise on SAR images, Remote Sens. Environ., China, vol. 11, no. 3, pp , [26] M. Ciuc, P. Bolon, E. Trouvé, V. Buzuloiu, and J.-P. Rudant, Adaptiveneighborhood speckle removal in multitemporal aperture radar images, Appl. Opt., vol. 40, no. 32, pp , Nov [27] B. Waske, M. Braun, and G. Menz, A segment-based speckle filter using multisensoral remote sensing imagery, IEEE Geosci. Remote Sens. Lett., vol. 4, no. 2, pp , Apr [28] J. W. Goodman, Some fundamental properties of speckle, J. Opt. Soc. Amer., vol. 66, no. 11, pp , Nov [29] H. Maître, Processing of Synthetic Aperture Radar Images. London, U.K.: Wiley, 2008, pp. 108, 153. [30] C. Oliver and S. Quegan, Understanding Synthetic Aperture Radar Images. Boston, MA: Artech House, 1998, pp [31] J. S. Lee and K. Hoppel, Noise modeling and estimation of remotely sensed images, in Proc. IGARSS, Vancouver, BC, Canada, Jul. 1989, vol. 2, pp [32] S. Smith and J. Brady, Susan A new approach to low level image processing, Int. J. Comput. Vis., vol. 23, no. 1, pp , May [33] C. Tomasi and R. Manduchi, Bilateral filtering for gray and color images, in Proc. 6th Int. Conf. Comput. Vis., 1998, pp [34] R. Touzi, A. Lopes, and P. Bousquet, Statistical and geometrical edge detector for SAR images, IEEE Trans. Geosci. Remote Sens., vol. 26, no. 6, pp , Nov [35] H. Xie, L. E. Pierce, and F. T. Ulaby, SAR speckle reduction using wavelet denoising and Markov random field modeling, IEEE Trans. Geosci. Remote Sens., vol. 40, no. 11, pp , Oct [36] J. S. Lee, K. Hoppel, and S. A. Mango, Unsupervised estimation of speckle noise in radar images, Int. J. Imag. Syst. Technol.,vol.4,pp , Hongxiao Feng (S 10) received the B.S. degree from Xidian University, Xian, China, in 2006, where he is currently working toward the Ph.D. degree at the Key Laboratory of Intelligent Perception and Image Understanding of the Ministry of Education of China. His current interests include synthetic aperture radar image interpretation, optical-satellite image processing, and the application of wavelet analysis in image processing. Biao Hou (M 07) was born in China, in He received the B.S. and M.S. degrees in mathematics from Northwest University, Xi an, China, in 1996 and 1999, respectively, and the Ph.D. degree in circuits and systems from Xidian University, Xi an, in Since 2003, he has been with the Key Laboratory of Intelligent Perception and Image Understanding of the Ministry of Education, Xidian University, where he is currently a Professor. His research interests include multiscale geometric analysis and synthetic aperture radar image processing. Maoguo Gong (M 07) received the B.Eng. degree with first class honors in electronic engineering and the Ph.D. degree from Xidian University, Xi an, China, in 2003 and 2009, respectively. He is currently with the Key Laboratory of Intelligent Perception and Image Understanding of the Ministry of Education of China, Xidian University, where he was promoted to Associate Professor and Full Professor in 2008 and 2010, respectively, both with exceptive admission. His research interests are broadly in the areas of computational intelligence and hybrid intelligent systems, with applications to optimization, data mining, and image understanding. He has published over 70 papers in journals and conferences, and he is the holder of eight patents. Dr. Gong is a member of the IEEE Computational Intelligence Society, executive committee member of the Natural Computation Society of Chinese Association for Artificial Intelligence, and senior member of the Chinese Computer Federation. He was the recipient of the New Century Excellent Talent in University of the Ministry of Education of China, the Eighth Young Scientist Award of Shaanxi, the New Scientific and Technological Star of Shaanxi Province, the Excellent Yong Contributor of Shaanxi Province, the Elsevier Scopus Young Scientist Award in Sustainable Development of China, and the Science and Technology Award of Shaanxi Province (First Level, 2008 and 2010), etc.

A Novel Adaptive Stationary Wavelet-based Technique for SAR Image Despeckling

A Novel Adaptive Stationary Wavelet-based Technique for SAR Image Despeckling A Novel Adaptive Stationary Wavelet-based Technique for SAR Image Despecling Amlan Jyoti Das Dept. of Electronics & ommunication Engineering, Gauhati University, Guwahati- 14, Assam Anjan Kumar Taludar

More information

MULTI-TEMPORAL SAR CHANGE DETECTION AND MONITORING

MULTI-TEMPORAL SAR CHANGE DETECTION AND MONITORING MULTI-TEMPORAL SAR CHANGE DETECTION AND MONITORING S. Hachicha, F. Chaabane Carthage University, Sup Com, COSIM laboratory, Route de Raoued, 3.5 Km, Elghazala Tunisia. ferdaous.chaabene@supcom.rnu.tn KEY

More information

Structure-adaptive Image Denoising with 3D Collaborative Filtering

Structure-adaptive Image Denoising with 3D Collaborative Filtering , pp.42-47 http://dx.doi.org/10.14257/astl.2015.80.09 Structure-adaptive Image Denoising with 3D Collaborative Filtering Xuemei Wang 1, Dengyin Zhang 2, Min Zhu 2,3, Yingtian Ji 2, Jin Wang 4 1 College

More information

IMAGE RESTORATION VIA EFFICIENT GAUSSIAN MIXTURE MODEL LEARNING

IMAGE RESTORATION VIA EFFICIENT GAUSSIAN MIXTURE MODEL LEARNING IMAGE RESTORATION VIA EFFICIENT GAUSSIAN MIXTURE MODEL LEARNING Jianzhou Feng Li Song Xiaog Huo Xiaokang Yang Wenjun Zhang Shanghai Digital Media Processing Transmission Key Lab, Shanghai Jiaotong University

More information

Compression of RADARSAT Data with Block Adaptive Wavelets Abstract: 1. Introduction

Compression of RADARSAT Data with Block Adaptive Wavelets Abstract: 1. Introduction Compression of RADARSAT Data with Block Adaptive Wavelets Ian Cumming and Jing Wang Department of Electrical and Computer Engineering The University of British Columbia 2356 Main Mall, Vancouver, BC, Canada

More information

Adaptive Doppler centroid estimation algorithm of airborne SAR

Adaptive Doppler centroid estimation algorithm of airborne SAR Adaptive Doppler centroid estimation algorithm of airborne SAR Jian Yang 1,2a), Chang Liu 1, and Yanfei Wang 1 1 Institute of Electronics, Chinese Academy of Sciences 19 North Sihuan Road, Haidian, Beijing

More information

Image Denoising Based on Hybrid Fourier and Neighborhood Wavelet Coefficients Jun Cheng, Songli Lei

Image Denoising Based on Hybrid Fourier and Neighborhood Wavelet Coefficients Jun Cheng, Songli Lei Image Denoising Based on Hybrid Fourier and Neighborhood Wavelet Coefficients Jun Cheng, Songli Lei College of Physical and Information Science, Hunan Normal University, Changsha, China Hunan Art Professional

More information

DUE to the acquisition under all weather conditions and

DUE to the acquisition under all weather conditions and IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 52, NO. 9, SEPTEMBER 2014 5751 Local Maximal Homogeneous Region Search for SAR Speckle Reduction With Sketch-Based Geometrical Kernel Function Jie

More information

Image Quality Assessment based on Improved Structural SIMilarity

Image Quality Assessment based on Improved Structural SIMilarity Image Quality Assessment based on Improved Structural SIMilarity Jinjian Wu 1, Fei Qi 2, and Guangming Shi 3 School of Electronic Engineering, Xidian University, Xi an, Shaanxi, 710071, P.R. China 1 jinjian.wu@mail.xidian.edu.cn

More information

Improved Non-Local Means Algorithm Based on Dimensionality Reduction

Improved Non-Local Means Algorithm Based on Dimensionality Reduction Improved Non-Local Means Algorithm Based on Dimensionality Reduction Golam M. Maruf and Mahmoud R. El-Sakka (&) Department of Computer Science, University of Western Ontario, London, Ontario, Canada {gmaruf,melsakka}@uwo.ca

More information

An Effective Denoising Method for Images Contaminated with Mixed Noise Based on Adaptive Median Filtering and Wavelet Threshold Denoising

An Effective Denoising Method for Images Contaminated with Mixed Noise Based on Adaptive Median Filtering and Wavelet Threshold Denoising J Inf Process Syst, Vol.14, No.2, pp.539~551, April 2018 https://doi.org/10.3745/jips.02.0083 ISSN 1976-913X (Print) ISSN 2092-805X (Electronic) An Effective Denoising Method for Images Contaminated with

More information

AN ALGORITHM FOR BLIND RESTORATION OF BLURRED AND NOISY IMAGES

AN ALGORITHM FOR BLIND RESTORATION OF BLURRED AND NOISY IMAGES AN ALGORITHM FOR BLIND RESTORATION OF BLURRED AND NOISY IMAGES Nader Moayeri and Konstantinos Konstantinides Hewlett-Packard Laboratories 1501 Page Mill Road Palo Alto, CA 94304-1120 moayeri,konstant@hpl.hp.com

More information

Image Denoising Methods Based on Wavelet Transform and Threshold Functions

Image Denoising Methods Based on Wavelet Transform and Threshold Functions Image Denoising Methods Based on Wavelet Transform and Threshold Functions Liangang Feng, Lin Lin Weihai Vocational College China liangangfeng@163.com liangangfeng@163.com ABSTRACT: There are many unavoidable

More information

Iterative Removing Salt and Pepper Noise based on Neighbourhood Information

Iterative Removing Salt and Pepper Noise based on Neighbourhood Information Iterative Removing Salt and Pepper Noise based on Neighbourhood Information Liu Chun College of Computer Science and Information Technology Daqing Normal University Daqing, China Sun Bishen Twenty-seventh

More information

IMAGE DE-NOISING IN WAVELET DOMAIN

IMAGE DE-NOISING IN WAVELET DOMAIN IMAGE DE-NOISING IN WAVELET DOMAIN Aaditya Verma a, Shrey Agarwal a a Department of Civil Engineering, Indian Institute of Technology, Kanpur, India - (aaditya, ashrey)@iitk.ac.in KEY WORDS: Wavelets,

More information

Recognition of Changes in SAR Images Based on Gauss-Log Ratio and MRFFCM

Recognition of Changes in SAR Images Based on Gauss-Log Ratio and MRFFCM Recognition of Changes in SAR Images Based on Gauss-Log Ratio and MRFFCM Jismy Alphonse M.Tech Scholar Computer Science and Engineering Department College of Engineering Munnar, Kerala, India Biju V. G.

More information

Patch-Based Color Image Denoising using efficient Pixel-Wise Weighting Techniques

Patch-Based Color Image Denoising using efficient Pixel-Wise Weighting Techniques Patch-Based Color Image Denoising using efficient Pixel-Wise Weighting Techniques Syed Gilani Pasha Assistant Professor, Dept. of ECE, School of Engineering, Central University of Karnataka, Gulbarga,

More information

AN INSITU SINGLE-POINTED WAVELET-BASED METHOD FOR NOISE REDUCTION IN SAR IMAGES

AN INSITU SINGLE-POINTED WAVELET-BASED METHOD FOR NOISE REDUCTION IN SAR IMAGES AN INSITU SINGLE-POINTED WAVELET-BASED METHOD FOR NOISE REDUCTION IN SAR IMAGES a, Huan Gu *, Guo Zhang a, Jun Yan a a State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing,

More information

Image Quality Assessment Techniques: An Overview

Image Quality Assessment Techniques: An Overview Image Quality Assessment Techniques: An Overview Shruti Sonawane A. M. Deshpande Department of E&TC Department of E&TC TSSM s BSCOER, Pune, TSSM s BSCOER, Pune, Pune University, Maharashtra, India Pune

More information

Learning based face hallucination techniques: A survey

Learning based face hallucination techniques: A survey Vol. 3 (2014-15) pp. 37-45. : A survey Premitha Premnath K Department of Computer Science & Engineering Vidya Academy of Science & Technology Thrissur - 680501, Kerala, India (email: premithakpnath@gmail.com)

More information

SAR change detection based on Generalized Gamma distribution. divergence and auto-threshold segmentation

SAR change detection based on Generalized Gamma distribution. divergence and auto-threshold segmentation SAR change detection based on Generalized Gamma distribution divergence and auto-threshold segmentation GAO Cong-shan 1 2, ZHANG Hong 1*, WANG Chao 1 1.Center for Earth Observation and Digital Earth, CAS,

More information

Keywords Change detection, Erosion, Morphological processing, Similarity measure, Spherically invariant random vector (SIRV) distribution models.

Keywords Change detection, Erosion, Morphological processing, Similarity measure, Spherically invariant random vector (SIRV) distribution models. Volume 5, Issue 3, March 2015 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Change Detection

More information

Image Denoising based on Spatial/Wavelet Filter using Hybrid Thresholding Function

Image Denoising based on Spatial/Wavelet Filter using Hybrid Thresholding Function Image Denoising based on Spatial/Wavelet Filter using Hybrid Thresholding Function Sabahaldin A. Hussain Electrical & Electronic Eng. Department University of Omdurman Sudan Sami M. Gorashi Electrical

More information

Markov Random Fields and Gibbs Sampling for Image Denoising

Markov Random Fields and Gibbs Sampling for Image Denoising Markov Random Fields and Gibbs Sampling for Image Denoising Chang Yue Electrical Engineering Stanford University changyue@stanfoed.edu Abstract This project applies Gibbs Sampling based on different Markov

More information

A Fast Speckle Reduction Algorithm based on GPU for Synthetic Aperture Sonar

A Fast Speckle Reduction Algorithm based on GPU for Synthetic Aperture Sonar Vol.137 (SUComS 016), pp.8-17 http://dx.doi.org/1457/astl.016.137.0 A Fast Speckle Reduction Algorithm based on GPU for Synthetic Aperture Sonar Xu Kui 1, Zhong Heping 1, Huang Pan 1 1 Naval Institute

More information

Novel speed up strategies for NLM Denoising With Patch Based Dictionaries

Novel speed up strategies for NLM Denoising With Patch Based Dictionaries Novel speed up strategies for NLM Denoising With Patch Based Dictionaries Rachita Shrivastava,Mrs Varsha Namdeo, Dr.Tripti Arjariya Abstract In this paper, a novel technique to speed-up a nonlocal means

More information

Image denoising using curvelet transform: an approach for edge preservation

Image denoising using curvelet transform: an approach for edge preservation Journal of Scientific & Industrial Research Vol. 3469, January 00, pp. 34-38 J SCI IN RES VOL 69 JANUARY 00 Image denoising using curvelet transform: an approach for edge preservation Anil A Patil * and

More information

IMAGE DENOISING USING FRAMELET TRANSFORM

IMAGE DENOISING USING FRAMELET TRANSFORM IMAGE DENOISING USING FRAMELET TRANSFORM Ms. Jadhav P.B. 1, Dr.Sangale.S.M. 2 1,2, Electronics Department,Shivaji University, (India) ABSTRACT Images are created to record or display useful information

More information

Stripe Noise Removal from Remote Sensing Images Based on Stationary Wavelet Transform

Stripe Noise Removal from Remote Sensing Images Based on Stationary Wavelet Transform Sensors & Transducers, Vol. 78, Issue 9, September 204, pp. 76-8 Sensors & Transducers 204 by IFSA Publishing, S. L. http://www.sensorsportal.com Stripe Noise Removal from Remote Sensing Images Based on

More information

Research on the New Image De-Noising Methodology Based on Neural Network and HMM-Hidden Markov Models

Research on the New Image De-Noising Methodology Based on Neural Network and HMM-Hidden Markov Models Research on the New Image De-Noising Methodology Based on Neural Network and HMM-Hidden Markov Models Wenzhun Huang 1, a and Xinxin Xie 1, b 1 School of Information Engineering, Xijing University, Xi an

More information

Image denoising in the wavelet domain using Improved Neigh-shrink

Image denoising in the wavelet domain using Improved Neigh-shrink Image denoising in the wavelet domain using Improved Neigh-shrink Rahim Kamran 1, Mehdi Nasri, Hossein Nezamabadi-pour 3, Saeid Saryazdi 4 1 Rahimkamran008@gmail.com nasri_me@yahoo.com 3 nezam@uk.ac.ir

More information

MULTICHANNEL image processing is studied in this

MULTICHANNEL image processing is studied in this 186 IEEE SIGNAL PROCESSING LETTERS, VOL. 6, NO. 7, JULY 1999 Vector Median-Rational Hybrid Filters for Multichannel Image Processing Lazhar Khriji and Moncef Gabbouj, Senior Member, IEEE Abstract In this

More information

Filters. Advanced and Special Topics: Filters. Filters

Filters. Advanced and Special Topics: Filters. Filters Filters Advanced and Special Topics: Filters Dr. Edmund Lam Department of Electrical and Electronic Engineering The University of Hong Kong ELEC4245: Digital Image Processing (Second Semester, 2016 17)

More information

Filtering Images. Contents

Filtering Images. Contents Image Processing and Data Visualization with MATLAB Filtering Images Hansrudi Noser June 8-9, 010 UZH, Multimedia and Robotics Summer School Noise Smoothing Filters Sigmoid Filters Gradient Filters Contents

More information

NSCT domain image fusion, denoising & K-means clustering for SAR image change detection

NSCT domain image fusion, denoising & K-means clustering for SAR image change detection NSCT domain image fusion, denoising & K-means clustering for SAR image change detection Yamuna J. 1, Arathy C. Haran 2 1,2, Department of Electronics and Communications Engineering, 1 P. G. student, 2

More information

Title. Author(s)Smolka, Bogdan. Issue Date Doc URL. Type. Note. File Information. Ranked-Based Vector Median Filter

Title. Author(s)Smolka, Bogdan. Issue Date Doc URL. Type. Note. File Information. Ranked-Based Vector Median Filter Title Ranked-Based Vector Median Filter Author(s)Smolka, Bogdan Proceedings : APSIPA ASC 2009 : Asia-Pacific Signal Citationand Conference: 254-257 Issue Date 2009-10-04 Doc URL http://hdl.handle.net/2115/39685

More information

Robust Shape Retrieval Using Maximum Likelihood Theory

Robust Shape Retrieval Using Maximum Likelihood Theory Robust Shape Retrieval Using Maximum Likelihood Theory Naif Alajlan 1, Paul Fieguth 2, and Mohamed Kamel 1 1 PAMI Lab, E & CE Dept., UW, Waterloo, ON, N2L 3G1, Canada. naif, mkamel@pami.uwaterloo.ca 2

More information

Engineering And Technology (affiliated to Anna University, Chennai) Tamil. Nadu, India

Engineering And Technology (affiliated to Anna University, Chennai) Tamil. Nadu, India International Journal of Advances in Engineering & Scientific Research, Vol.2, Issue 2, Feb - 2015, pp 08-13 ISSN: 2349 3607 (Online), ISSN: 2349 4824 (Print) ABSTRACT MULTI-TEMPORAL SAR IMAGE CHANGE DETECTION

More information

MR IMAGE SEGMENTATION

MR IMAGE SEGMENTATION MR IMAGE SEGMENTATION Prepared by : Monil Shah What is Segmentation? Partitioning a region or regions of interest in images such that each region corresponds to one or more anatomic structures Classification

More information

A reversible data hiding based on adaptive prediction technique and histogram shifting

A reversible data hiding based on adaptive prediction technique and histogram shifting A reversible data hiding based on adaptive prediction technique and histogram shifting Rui Liu, Rongrong Ni, Yao Zhao Institute of Information Science Beijing Jiaotong University E-mail: rrni@bjtu.edu.cn

More information

WAVELET DENOISING WITH EDGE DETECTION FOR SPECKLE REDUCTION IN SAR IMAGES

WAVELET DENOISING WITH EDGE DETECTION FOR SPECKLE REDUCTION IN SAR IMAGES WAVELET DENOISING WITH EDGE DETECTION FOR SPECKLE REDUCTION IN SAR IMAGES M. Rosa-Zurera, A.M. Cóbreces-Álvarez, J.C. Nieto-Borge, M.P. Jarabo-Amores, and D. Mata-Moya Departamento de Teoría de la Señal

More information

Hyperspectral and Multispectral Image Fusion Using Local Spatial-Spectral Dictionary Pair

Hyperspectral and Multispectral Image Fusion Using Local Spatial-Spectral Dictionary Pair Hyperspectral and Multispectral Image Fusion Using Local Spatial-Spectral Dictionary Pair Yifan Zhang, Tuo Zhao, and Mingyi He School of Electronics and Information International Center for Information

More information

Image Enhancement Techniques for Fingerprint Identification

Image Enhancement Techniques for Fingerprint Identification March 2013 1 Image Enhancement Techniques for Fingerprint Identification Pankaj Deshmukh, Siraj Pathan, Riyaz Pathan Abstract The aim of this paper is to propose a new method in fingerprint enhancement

More information

An Optimized Pixel-Wise Weighting Approach For Patch-Based Image Denoising

An Optimized Pixel-Wise Weighting Approach For Patch-Based Image Denoising An Optimized Pixel-Wise Weighting Approach For Patch-Based Image Denoising Dr. B. R.VIKRAM M.E.,Ph.D.,MIEEE.,LMISTE, Principal of Vijay Rural Engineering College, NIZAMABAD ( Dt.) G. Chaitanya M.Tech,

More information

Exact discrete minimization for TV+L0 image decomposition models

Exact discrete minimization for TV+L0 image decomposition models Exact discrete minimization for TV+L0 image decomposition models Loïc Denis 1, Florence Tupin 2 and Xavier Rondeau 2 1. Observatory of Lyon (CNRS / Univ. Lyon 1 / ENS de Lyon), France 2. Telecom ParisTech

More information

Denoising an Image by Denoising its Components in a Moving Frame

Denoising an Image by Denoising its Components in a Moving Frame Denoising an Image by Denoising its Components in a Moving Frame Gabriela Ghimpețeanu 1, Thomas Batard 1, Marcelo Bertalmío 1, and Stacey Levine 2 1 Universitat Pompeu Fabra, Spain 2 Duquesne University,

More information

Region Based Image Fusion Using SVM

Region Based Image Fusion Using SVM Region Based Image Fusion Using SVM Yang Liu, Jian Cheng, Hanqing Lu National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences ABSTRACT This paper presents a novel

More information

Speckle Noise Removal Using Dual Tree Complex Wavelet Transform

Speckle Noise Removal Using Dual Tree Complex Wavelet Transform Speckle Noise Removal Using Dual Tree Complex Wavelet Transform Dr. Siva Agora Sakthivel Murugan, K.Karthikayan, Natraj.N.A, Rathish.C.R Abstract: Dual Tree Complex Wavelet Transform (DTCWT),is a form

More information

PERFORMANCE EVALUATION OF ADAPTIVE SPECKLE FILTERS FOR ULTRASOUND IMAGES

PERFORMANCE EVALUATION OF ADAPTIVE SPECKLE FILTERS FOR ULTRASOUND IMAGES PERFORMANCE EVALUATION OF ADAPTIVE SPECKLE FILTERS FOR ULTRASOUND IMAGES Abstract: L.M.Merlin Livingston #, L.G.X.Agnel Livingston *, Dr. L.M.Jenila Livingston ** #Associate Professor, ECE Dept., Jeppiaar

More information

Evaluation of texture features for image segmentation

Evaluation of texture features for image segmentation RIT Scholar Works Articles 9-14-2001 Evaluation of texture features for image segmentation Navid Serrano Jiebo Luo Andreas Savakis Follow this and additional works at: http://scholarworks.rit.edu/article

More information

Fast and Effective Interpolation Using Median Filter

Fast and Effective Interpolation Using Median Filter Fast and Effective Interpolation Using Median Filter Jian Zhang 1, *, Siwei Ma 2, Yongbing Zhang 1, and Debin Zhao 1 1 Department of Computer Science, Harbin Institute of Technology, Harbin 150001, P.R.

More information

RESTORATION OF TEXTURAL PROPERTIES IN SAR IMAGES USING SECOND ORDER STATISTICS

RESTORATION OF TEXTURAL PROPERTIES IN SAR IMAGES USING SECOND ORDER STATISTICS RESTORATION OF TEXTURAL PROPERTIES IN SAR IMAGES USING SECOND ORDER STATISTICS Edmond NEZRY, Hans-Günter KOHL, Hugo DE GROOF Joint Research Centre of the European Communities (JRC) Institute for Remote

More information

IMAGE DENOISING USING NL-MEANS VIA SMOOTH PATCH ORDERING

IMAGE DENOISING USING NL-MEANS VIA SMOOTH PATCH ORDERING IMAGE DENOISING USING NL-MEANS VIA SMOOTH PATCH ORDERING Idan Ram, Michael Elad and Israel Cohen Department of Electrical Engineering Department of Computer Science Technion - Israel Institute of Technology

More information

CHAPTER 2 TEXTURE CLASSIFICATION METHODS GRAY LEVEL CO-OCCURRENCE MATRIX AND TEXTURE UNIT

CHAPTER 2 TEXTURE CLASSIFICATION METHODS GRAY LEVEL CO-OCCURRENCE MATRIX AND TEXTURE UNIT CHAPTER 2 TEXTURE CLASSIFICATION METHODS GRAY LEVEL CO-OCCURRENCE MATRIX AND TEXTURE UNIT 2.1 BRIEF OUTLINE The classification of digital imagery is to extract useful thematic information which is one

More information

IMAGE DENOISING TO ESTIMATE THE GRADIENT HISTOGRAM PRESERVATION USING VARIOUS ALGORITHMS

IMAGE DENOISING TO ESTIMATE THE GRADIENT HISTOGRAM PRESERVATION USING VARIOUS ALGORITHMS IMAGE DENOISING TO ESTIMATE THE GRADIENT HISTOGRAM PRESERVATION USING VARIOUS ALGORITHMS P.Mahalakshmi 1, J.Muthulakshmi 2, S.Kannadhasan 3 1,2 U.G Student, 3 Assistant Professor, Department of Electronics

More information

Modified Directional Weighted Median Filter

Modified Directional Weighted Median Filter Modified Directional Weighted Median Filter Ayyaz Hussain 1, Muhammad Asim Khan 2, Zia Ul-Qayyum 2 1 Faculty of Basic and Applied Sciences, Department of Computer Science, Islamic International University

More information

A MORPHOLOGY-BASED FILTER STRUCTURE FOR EDGE-ENHANCING SMOOTHING

A MORPHOLOGY-BASED FILTER STRUCTURE FOR EDGE-ENHANCING SMOOTHING Proceedings of the 1994 IEEE International Conference on Image Processing (ICIP-94), pp. 530-534. (Austin, Texas, 13-16 November 1994.) A MORPHOLOGY-BASED FILTER STRUCTURE FOR EDGE-ENHANCING SMOOTHING

More information

SAR Speckle Filtering

SAR Speckle Filtering SAR Speckle Filtering SAR Training for forest monitoring 014/015 Cédric Lardeux Jean-Paul Rudant Pierre-Louis Frison cedric.lardeux@onfinternational.com rudant@univ-mlv.fr frison@univ-mlv.fr SAR for Forest

More information

A Novel Extreme Point Selection Algorithm in SIFT

A Novel Extreme Point Selection Algorithm in SIFT A Novel Extreme Point Selection Algorithm in SIFT Ding Zuchun School of Electronic and Communication, South China University of Technolog Guangzhou, China zucding@gmail.com Abstract. This paper proposes

More information

A Simple Algorithm for Image Denoising Based on MS Segmentation

A Simple Algorithm for Image Denoising Based on MS Segmentation A Simple Algorithm for Image Denoising Based on MS Segmentation G.Vijaya 1 Dr.V.Vasudevan 2 Senior Lecturer, Dept. of CSE, Kalasalingam University, Krishnankoil, Tamilnadu, India. Senior Prof. & Head Dept.

More information

Texture Image Segmentation using FCM

Texture Image Segmentation using FCM Proceedings of 2012 4th International Conference on Machine Learning and Computing IPCSIT vol. 25 (2012) (2012) IACSIT Press, Singapore Texture Image Segmentation using FCM Kanchan S. Deshmukh + M.G.M

More information

Image Denoising using SWT 2D Wavelet Transform

Image Denoising using SWT 2D Wavelet Transform IJSTE - International Journal of Science Technology & Engineering Volume 3 Issue 01 July 2016 ISSN (online): 2349-784X Image Denoising using SWT 2D Wavelet Transform Deep Singh Bedi Department of Electronics

More information

Express Letters. A Simple and Efficient Search Algorithm for Block-Matching Motion Estimation. Jianhua Lu and Ming L. Liou

Express Letters. A Simple and Efficient Search Algorithm for Block-Matching Motion Estimation. Jianhua Lu and Ming L. Liou IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 7, NO. 2, APRIL 1997 429 Express Letters A Simple and Efficient Search Algorithm for Block-Matching Motion Estimation Jianhua Lu and

More information

Research on the Image Denoising Method Based on Partial Differential Equations

Research on the Image Denoising Method Based on Partial Differential Equations BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 5 Special Issue on Application of Advanced Computing and Simulation in Information Systems Sofia 2016 Print ISSN: 1311-9702;

More information

Journal of Asian Scientific Research FEATURES COMPOSITION FOR PROFICIENT AND REAL TIME RETRIEVAL IN CBIR SYSTEM. Tohid Sedghi

Journal of Asian Scientific Research FEATURES COMPOSITION FOR PROFICIENT AND REAL TIME RETRIEVAL IN CBIR SYSTEM. Tohid Sedghi Journal of Asian Scientific Research, 013, 3(1):68-74 Journal of Asian Scientific Research journal homepage: http://aessweb.com/journal-detail.php?id=5003 FEATURES COMPOSTON FOR PROFCENT AND REAL TME RETREVAL

More information

A Fourier Extension Based Algorithm for Impulse Noise Removal

A Fourier Extension Based Algorithm for Impulse Noise Removal A Fourier Extension Based Algorithm for Impulse Noise Removal H. Sahoolizadeh, R. Rajabioun *, M. Zeinali Abstract In this paper a novel Fourier extension based algorithm is introduced which is able to

More information

PRINCIPAL COMPONENT ANALYSIS IMAGE DENOISING USING LOCAL PIXEL GROUPING

PRINCIPAL COMPONENT ANALYSIS IMAGE DENOISING USING LOCAL PIXEL GROUPING PRINCIPAL COMPONENT ANALYSIS IMAGE DENOISING USING LOCAL PIXEL GROUPING Divesh Kumar 1 and Dheeraj Kalra 2 1 Department of Electronics & Communication Engineering, IET, GLA University, Mathura 2 Department

More information

Robust Image Watermarking based on DCT-DWT- SVD Method

Robust Image Watermarking based on DCT-DWT- SVD Method Robust Image Watermarking based on DCT-DWT- SVD Sneha Jose Rajesh Cherian Roy, PhD. Sreenesh Shashidharan ABSTRACT Hybrid Image watermarking scheme proposed based on Discrete Cosine Transform (DCT)-Discrete

More information

Defense Technology, Changsha , P. R. China

Defense Technology, Changsha , P. R. China Progress In Electromagnetics Research M, Vol. 18, 259 269, 2011 SAR IMAGE MATCHING METHOD BASED ON IMPROVED SIFT FOR NAVIGATION SYSTEM S. Ren 1, *, W. Chang 1, and X. Liu 2 1 School of Electronic Science

More information

A WAVELET DOMAIN FILTER FOR CORRELATED SPECKLE

A WAVELET DOMAIN FILTER FOR CORRELATED SPECKLE A WAVELET DOMAIN FILTER FOR CORRELATED SPECKLE Stian Solbø and Torbjørn Eltoft Norut IT, Tromsø, Norway. Tel: +47 776 9 45, e-mail: stian.solboe@itek,norut,no Institute of Physics, University of Tromsø,

More information

Quaternion-based color difference measure for removing impulse noise in color images

Quaternion-based color difference measure for removing impulse noise in color images 2014 International Conference on Informative and Cybernetics for Computational Social Systems (ICCSS) Quaternion-based color difference measure for removing impulse noise in color images Lunbo Chen, Yicong

More information

DESIGN OF A NOVEL IMAGE FUSION ALGORITHM FOR IMPULSE NOISE REMOVAL IN REMOTE SENSING IMAGES BY USING THE QUALITY ASSESSMENT

DESIGN OF A NOVEL IMAGE FUSION ALGORITHM FOR IMPULSE NOISE REMOVAL IN REMOTE SENSING IMAGES BY USING THE QUALITY ASSESSMENT DESIGN OF A NOVEL IMAGE FUSION ALGORITHM FOR IMPULSE NOISE REMOVAL IN REMOTE SENSING IMAGES BY USING THE QUALITY ASSESSMENT P.PAVANI, M.V.H.BHASKARA MURTHY Department of Electronics and Communication Engineering,Aditya

More information

IMPLEMENTATION OF THE CONTRAST ENHANCEMENT AND WEIGHTED GUIDED IMAGE FILTERING ALGORITHM FOR EDGE PRESERVATION FOR BETTER PERCEPTION

IMPLEMENTATION OF THE CONTRAST ENHANCEMENT AND WEIGHTED GUIDED IMAGE FILTERING ALGORITHM FOR EDGE PRESERVATION FOR BETTER PERCEPTION IMPLEMENTATION OF THE CONTRAST ENHANCEMENT AND WEIGHTED GUIDED IMAGE FILTERING ALGORITHM FOR EDGE PRESERVATION FOR BETTER PERCEPTION Chiruvella Suresh Assistant professor, Department of Electronics & Communication

More information

A Novel Algorithm for Color Image matching using Wavelet-SIFT

A Novel Algorithm for Color Image matching using Wavelet-SIFT International Journal of Scientific and Research Publications, Volume 5, Issue 1, January 2015 1 A Novel Algorithm for Color Image matching using Wavelet-SIFT Mupuri Prasanth Babu *, P. Ravi Shankar **

More information

Speckle Suppression of Radar Images Using Normalized Convolution

Speckle Suppression of Radar Images Using Normalized Convolution Journal of Computer Science 6 (10): 1154-1158, 2010 ISSN 1549-3636 2010 Science Publications Speckle Suppression of Radar Images Using Normalized Convolution 1 A.K. Helmy and 2 G.S. El-Taweel 1 National

More information

DENOISING SONAR IMAGES USING A BISHRINK FILTER WITH REDUCED SENSITIVITY

DENOISING SONAR IMAGES USING A BISHRINK FILTER WITH REDUCED SENSITIVITY Électronique et transmission de l information DENOISING SONAR IMAGES USING A BISHRINK FILTER WITH REDUCED SENSITIVITY ALEXANDRU ISAR 1, SORIN MOGA 2, DORINA ISAR 1 Key words: SONAR, MAP-filter, Double

More information

DYNAMIC BACKGROUND SUBTRACTION BASED ON SPATIAL EXTENDED CENTER-SYMMETRIC LOCAL BINARY PATTERN. Gengjian Xue, Jun Sun, Li Song

DYNAMIC BACKGROUND SUBTRACTION BASED ON SPATIAL EXTENDED CENTER-SYMMETRIC LOCAL BINARY PATTERN. Gengjian Xue, Jun Sun, Li Song DYNAMIC BACKGROUND SUBTRACTION BASED ON SPATIAL EXTENDED CENTER-SYMMETRIC LOCAL BINARY PATTERN Gengjian Xue, Jun Sun, Li Song Institute of Image Communication and Information Processing, Shanghai Jiao

More information

Resolution Magnification Technique for Satellite Images Using DT- CWT and NLM

Resolution Magnification Technique for Satellite Images Using DT- CWT and NLM AUSTRALIAN JOURNAL OF BASIC AND APPLIED SCIENCES ISSN:1991-8178 EISSN: 2309-8414 Journal home page: www.ajbasweb.com Resolution Magnification Technique for Satellite Images Using DT- CWT and NLM 1 Saranya

More information

SUPPLEMENTARY MATERIAL

SUPPLEMENTARY MATERIAL SUPPLEMENTARY MATERIAL Zhiyuan Zha 1,3, Xin Liu 2, Ziheng Zhou 2, Xiaohua Huang 2, Jingang Shi 2, Zhenhong Shang 3, Lan Tang 1, Yechao Bai 1, Qiong Wang 1, Xinggan Zhang 1 1 School of Electronic Science

More information

Blur Space Iterative De-blurring

Blur Space Iterative De-blurring Blur Space Iterative De-blurring RADU CIPRIAN BILCU 1, MEJDI TRIMECHE 2, SAKARI ALENIUS 3, MARKKU VEHVILAINEN 4 1,2,3,4 Multimedia Technologies Laboratory, Nokia Research Center Visiokatu 1, FIN-33720,

More information

Graph Matching Iris Image Blocks with Local Binary Pattern

Graph Matching Iris Image Blocks with Local Binary Pattern Graph Matching Iris Image Blocs with Local Binary Pattern Zhenan Sun, Tieniu Tan, and Xianchao Qiu Center for Biometrics and Security Research, National Laboratory of Pattern Recognition, Institute of

More information

Robust color segmentation algorithms in illumination variation conditions

Robust color segmentation algorithms in illumination variation conditions 286 CHINESE OPTICS LETTERS / Vol. 8, No. / March 10, 2010 Robust color segmentation algorithms in illumination variation conditions Jinhui Lan ( ) and Kai Shen ( Department of Measurement and Control Technologies,

More information

Non-local Means for Stereo Image Denoising Using Structural Similarity

Non-local Means for Stereo Image Denoising Using Structural Similarity Non-local Means for Stereo Image Denoising Using Structural Similarity Monagi H. Alkinani and Mahmoud R. El-Sakka (B) Computer Science Department, University of Western Ontario, London, ON N6A 5B7, Canada

More information

SHIP WAKE DETECTION FOR SAR IMAGES WITH COMPLEX BACKGROUNDS BASED ON MORPHOLOGICAL DICTIONARY LEARNING

SHIP WAKE DETECTION FOR SAR IMAGES WITH COMPLEX BACKGROUNDS BASED ON MORPHOLOGICAL DICTIONARY LEARNING SHIP WAKE DETECTION FOR SAR IMAGES WITH COMPLEX BACKGROUNDS BASED ON MORPHOLOGICAL DICTIONARY LEARNING Guozheng Yang 1, 2, Jing Yu 3, Chuangbai Xiao 3, Weidong Sun 1 1 State Key Laboratory of Intelligent

More information

MRF-based Algorithms for Segmentation of SAR Images

MRF-based Algorithms for Segmentation of SAR Images This paper originally appeared in the Proceedings of the 998 International Conference on Image Processing, v. 3, pp. 770-774, IEEE, Chicago, (998) MRF-based Algorithms for Segmentation of SAR Images Robert

More information

MULTI-TEMPORAL SAR DATA FILTERING FOR LAND APPLICATIONS. I i is the estimate of the local mean backscattering

MULTI-TEMPORAL SAR DATA FILTERING FOR LAND APPLICATIONS. I i is the estimate of the local mean backscattering MULTI-TEMPORAL SAR DATA FILTERING FOR LAND APPLICATIONS Urs Wegmüller (1), Maurizio Santoro (1), and Charles Werner (1) (1) Gamma Remote Sensing AG, Worbstrasse 225, CH-3073 Gümligen, Switzerland http://www.gamma-rs.ch,

More information

Advance Shadow Edge Detection and Removal (ASEDR)

Advance Shadow Edge Detection and Removal (ASEDR) International Journal of Computational Intelligence Research ISSN 0973-1873 Volume 13, Number 2 (2017), pp. 253-259 Research India Publications http://www.ripublication.com Advance Shadow Edge Detection

More information

Statistical image models

Statistical image models Chapter 4 Statistical image models 4. Introduction 4.. Visual worlds Figure 4. shows images that belong to different visual worlds. The first world (fig. 4..a) is the world of white noise. It is the world

More information

New structural similarity measure for image comparison

New structural similarity measure for image comparison University of Wollongong Research Online Faculty of Engineering and Information Sciences - Papers: Part A Faculty of Engineering and Information Sciences 2012 New structural similarity measure for image

More information

Image Denoising based on Adaptive BM3D and Singular Value

Image Denoising based on Adaptive BM3D and Singular Value Image Denoising based on Adaptive BM3D and Singular Value Decomposition YouSai hang, ShuJin hu, YuanJiang Li Institute of Electronic and Information, Jiangsu University of Science and Technology, henjiang,

More information

Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial Region Segmentation

Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial Region Segmentation IJCSNS International Journal of Computer Science and Network Security, VOL.13 No.11, November 2013 1 Moving Object Segmentation Method Based on Motion Information Classification by X-means and Spatial

More information

Adaptive Wavelet Image Denoising Based on the Entropy of Homogenus Regions

Adaptive Wavelet Image Denoising Based on the Entropy of Homogenus Regions International Journal of Electrical and Electronic Science 206; 3(4): 9-25 http://www.aascit.org/journal/ijees ISSN: 2375-2998 Adaptive Wavelet Image Denoising Based on the Entropy of Homogenus Regions

More information

COMPARISONS OF DCT-BASED AND DWT-BASED WATERMARKING TECHNIQUES

COMPARISONS OF DCT-BASED AND DWT-BASED WATERMARKING TECHNIQUES COMPARISONS OF DCT-BASED AND DWT-BASED WATERMARKING TECHNIQUES H. I. Saleh 1, M. E. Elhadedy 2, M. A. Ashour 1, M. A. Aboelsaud 3 1 Radiation Engineering Dept., NCRRT, AEA, Egypt. 2 Reactor Dept., NRC,

More information

A ROBUST LONE DIAGONAL SORTING ALGORITHM FOR DENOISING OF IMAGES WITH SALT AND PEPPER NOISE

A ROBUST LONE DIAGONAL SORTING ALGORITHM FOR DENOISING OF IMAGES WITH SALT AND PEPPER NOISE International Journal of Computational Intelligence & Telecommunication Systems, 2(1), 2011, pp. 33-38 A ROBUST LONE DIAGONAL SORTING ALGORITHM FOR DENOISING OF IMAGES WITH SALT AND PEPPER NOISE Rajamani.

More information

SSIM Image Quality Metric for Denoised Images

SSIM Image Quality Metric for Denoised Images SSIM Image Quality Metric for Denoised Images PETER NDAJAH, HISAKAZU KIKUCHI, MASAHIRO YUKAWA, HIDENORI WATANABE and SHOGO MURAMATSU Department of Electrical and Electronics Engineering, Niigata University,

More information

Change Detection in Remotely Sensed Images Based on Image Fusion and Fuzzy Clustering

Change Detection in Remotely Sensed Images Based on Image Fusion and Fuzzy Clustering International Journal of Electronics Engineering Research. ISSN 0975-6450 Volume 9, Number 1 (2017) pp. 141-150 Research India Publications http://www.ripublication.com Change Detection in Remotely Sensed

More information

An Approach for Reduction of Rain Streaks from a Single Image

An Approach for Reduction of Rain Streaks from a Single Image An Approach for Reduction of Rain Streaks from a Single Image Vijayakumar Majjagi 1, Netravati U M 2 1 4 th Semester, M. Tech, Digital Electronics, Department of Electronics and Communication G M Institute

More information

Unsupervised Oil Spill Detection in SAR Imagery through an Estimator of Local Regularity

Unsupervised Oil Spill Detection in SAR Imagery through an Estimator of Local Regularity Unsupervised Oil Spill Detection in SAR Imagery through an Estimator of Local Regularity Mariví Tello,, Carlos López-Martínez,, Jordi J. Mallorquí. Remote Sensing Laboratory (RSLab( RSLab) Signal Theory

More information

A NEW ROBUST IMAGE WATERMARKING SCHEME BASED ON DWT WITH SVD

A NEW ROBUST IMAGE WATERMARKING SCHEME BASED ON DWT WITH SVD A NEW ROBUST IMAGE WATERMARKING SCHEME BASED ON WITH S.Shanmugaprabha PG Scholar, Dept of Computer Science & Engineering VMKV Engineering College, Salem India N.Malmurugan Director Sri Ranganathar Institute

More information

Face Hallucination Based on Eigentransformation Learning

Face Hallucination Based on Eigentransformation Learning Advanced Science and Technology etters, pp.32-37 http://dx.doi.org/10.14257/astl.2016. Face allucination Based on Eigentransformation earning Guohua Zou School of software, East China University of Technology,

More information