International Journal of Computer Engineering and Applications, Volume XII, Issue I, Jan. 18, ISSN

Transcription

1 International Journal of Computer Engineering and Applications, Volume XII, Issue I, Jan. 18, ISSN AN EFFICIENT FEATURE BASED IMAGE SEGMENTATION OF SATELLITE IMAGES USING LOCAL THRESHOLD METHOD T. Balaji 1, K. Prabha 2 1 Assistant Professor, P.G. Dept. of Computer Science, Govt. Arts College, Melur, India 2 Post Graduate Student, P.G. Dept. of Computer Science, Govt. Arts College, Melur, India 1 bkmd_gacm@rediffmail.com 2 vani_bmd@rediffmail.com ABSTRACT A thresholding is a simple and well-known technique for satellite image segmentation. Because of its wide applicability to other areas of the digital image processing, quite a number of thresholding methods have been proposed over the years. In this paper, we present a segmentation approach of local thresholding techniques and update the histogram equalization. A popular tool used in image segmentation is thresholding approach. Satellite mage segmentation in remote sensing application is often used to distinguish the foreground from the background. This paper proposes a susceptible method of image thresholding using the optimal histogram segmentation in satellite image. Satellite image segmentation is very essential to image processing and pattern recognition. It leads to the high quality of the final result of analysis. Satellite image segmentation is a process of dividing an image into different regions or boundaries. One of the special kinds of segmentation is thresholding, which attempts to categorize image pixels into one of the two categories such as foreground and background. At the end of such thresholding, each object of the image, represented by a set of pixels, is isolated from the rest of the scene. We attempt to evaluate the performance of some automatic global thresholding methods using the criterion functions such as uniformity and shape measures. The evaluation is based on some real world images. This is the most straightforward approach is to pick up a fixed gray scale value as the threshold and categorize each gray scale by checking whether it lies above or below this value. In general, the threshold should be located at the obvious and deep valley of the histogram. Keywords - Automatic Threshloding, Hysteresis Thresholding, Optimal Thresholding, Variable Thresholding and Histogram. T. Balaji, K. Prabha 190

2 AN EFFICIENT FEATURE BASED IMAGE SEGMENTATION OF SATELLITE IMAGES USING LOCAL THRESHOLD METHOD I. INTRODUCTION Satellite image segmentation refers to the task of extracting information classes from a multiband raster image. The resulting raster from image segmentation can be used to create thematic maps. Depending on the interaction between the analyst and the computer during segmentation, there are two types of segmentation: supervised and unsupervised. The objective of image segmentation is to identify and portray, as a unique gray level (or color), the features occurring in an image in terms of the object or type of land cover these features actually represent on the ground. A broad group of digital image processing techniques of remote sensing application explains the image segmentation techniques are most generally applied to the spectral data of a single-date image or to the varying spectral data of a series of multidimensional images. Satellite based research focusing on image segmentation has long attracted the attention of the satellite community because segmentation results are the basis for many environmental and socioeconomic applications. Image segmentation is a complex process that may be affected by many factors. The different review methods are suggests that designing a suitable image processing procedure is a prerequisite for a successful segmentation of remotely sensed data into a thematic map. Effective use of multiple features of remotely sensed data and the selection of a suitable segmentation method are especially significant for improving segmentation accuracy. Continuous emergence of new segmentation algorithms and techniques in recent years necessitates such a review, which will be highly valuable for guiding or selecting a suitable segmentation procedure for a specific study. This paper focuses a summarization of local threshold segmentation methods and techniques used for improving segmentation accuracy. In many applications of image processing, the gray levels of pixels belonging to the object are quite different from the gray levels of the pixels belonging to the background. Thresholding becomes then a simple but effective tool to separate objects from the background. Examples of thresholding applications are document image analysis where the goal is to extract printed characters, logos, graphical content, musical scores, map processing where lines, legends, characters are to be found, scene processing where a target is to detected, quality inspection of materials. Other applications include cell images and knowledge representation, segmentation of various image modalities for nondestructive testing (NDT) applications, such as ultrasonic images in, eddy current images, thermal images, and x- ray, computed tomography, laser scanning confocal microscopy, extraction of edge field, image segmentation in general, spatio-temporal segmentation of video images etc. The output of the thresholding operation is a binary image whose gray level of 0 (black) will indicate a pixel belonging to a print, legend, drawing, or target and a gray level of 1 (white) will indicate the background. II. THRESHOLDING METHOD Thresholding is the simplest method of image segmentation. From a gray scale image, thresholding can be used to create binary images. Segmentation involves separating an image into regions (or their contours) corresponding to objects. We usually try to segment regions by identifying common properties. Similarly, we identify contours by identifying differences between regions (edges).the simplest property that pixels in a region can share is intensity. So, a natural way to segment such regions is through thresholding, the separation of light and dark regions. Thresholding creates binary images from grey-level ones by turning all pixels below some threshold to zero and all pixels about that threshold to one. If g(x, y) is a thresholded version of f(x, y) at some global threshold T, In image processing, thresholding is used to split an image into smaller segments, or junks, using at least one color or grayscale value to define their boundary. In that sense segments should represent objects in the image, e.g. letters, cars, traffic signs, human faces in the foreground. Unfortunately, such objects vary with respect to colors, intensity, illumination, lens aberrations and noise which make choosing the right threshold not an easy task. Multiple thresholds might be needed in connection with more sophisticated algorithms. Let N be the set of natural numbers, (x, y) be the spatial coordinate of a digitized image, and G L = (0, 1,..., l - 1) be a set of positive integers representing gray levels. Then, an image function can be defined as the mapping f: N x N -> G L. The brightness (i.e., gray level) of a pixel with coordinate (x, y) is denoted as f(x, y). Let T ϵ G L be a threshold and Bi = {b 0, b 1} be a pair of binary gray levels and b 0, b 1 ϵ G L. The result of thresholding an image function f(x, y) at gray level T is a binary image function f t : N x N -> B i, such that In general, a thresholding method is one that determines the value t* of t based on a certain criterion. If t* is determined solely from the gray level of each pixel, then the thresholding method is point-dependent. If t* is determined from the local property (e.g., the local gray level distribution) in the neighborhood of each pixel, then the thresholding method is region-dependent. A global thresholding technique is one that thresholds the entire image with a single threshold value, whereas a T. Balaji, K. Prabha 191

3 International Journal of Computer Engineering and Applications, Volume XII, Issue I, Jan. 18, ISSN local thresholding technique is one that partitions a given image into sub images and determines a threshold for each of these sub images. Let the number of pixels with gray level i be n i. Then the total number of pixels in a given image is The probability of occurrence of gray level i with respect to be Ni is defined as Also, by convention, the gray level 0 is the darkest and the gray level l - 1 is the lightest. III. DIFFERENT PRESPECTIVE OF THRESHOLDING We categorize the thresholding methods in six groups according to the information they are exploiting. These categories are: A. Histogram Shape-Based Method The histogram shape-based methods where the peaks, valleys and curvatures of the smoothed histogram are analyzed. This category of methods achieves thresholding based on the multilevel properties of the histogram. Basically two major peaks and an intervening valley is searched for using such tools as the convex hull of the histogram, or its curvature and zero crossings of the wavelet components. B. Clustering-Based Method Clustering-based methods, the gray level samples are clustered in two parts as background and foreground (object) or alternately are modeled as two Gaussian distributions. Understand that the place of minimum overlap (the place where the misclassified areas of the distributions are equal) is not necessarily where the valley occurs in the combined histogram. This occurs, when one cluster has a wide distribution and the other a narrow one. One way that we can try to do this is to consider the values in the two regions as two clusters. The idea is to pick a threshold such that each pixel on each side of the threshold is closer in intensity to the mean of all pixels on that side of the threshold than the mean of all pixels on the other side of the threshold. C. Entropy-Based Method Another class of thresholding method is entropybased techniques, where entropy is used to separate the global thresholding classes. For example, the optimal threshold value can be calculated by maximizing the sum of the foreground and background entropies, i.e. maximally separating region intensities of the foreground and background. Maximum entropy-based method used Shannon s concept to define the entropy of an image. This concept to derive an expression for an upper bound of the a posteriori entropy. The expression was finally used to find the threshold of an image. In this method has also has to be improved the histogram entropy based method. It is a histogram analysis and maximum-entropybased global technique, which uses the maximum of the sum of the entropy of the grey-level distribution of the foreground and background. The technique has shown good performance for picture images. It is also presented a higher order entropy method for object extraction and summarized several entropy methods used in image processing. The spatial context of thresholding algorithm that minimizes relative entropy. It uses spatial dependence (co-occurrence probability) of the pixels. This is applied a maximumrelative-entropy-based technique to picture images. D. Object Attribute-Based Method Object attribute-based methods search a measure of similarity between the gray-level and binarized images, such as fuzzy similarity, shape, edges, number of objects etc. The algorithms considered under this category select the threshold value based on some similarity measure between the original image and the binarized version of the image. These attributes can take the form of edges, shapes, or one can directly consider the original graylevel image to binary image resemblance. Alternately they consider certain image attributes such as compactness or connectivity of the objects resulting from the binarization process or the coincidence of the edge fields. E. Spatial Method The spatial methods use the probability mass function models taking into account correlation between pixels on a global scale. In this class of algorithms one utilizes spatial information of object and background pixels, for example, in the form of context probabilities, correlation functions, cooccurrence probabilities, local linear dependence models of pixels, two-dimensional entropy etc. One of the first to explore spatial information was considered such ideas as local average gray level for thresholding. Another spatial information used relaxation to improve on the binary map as in the Laplacian of the images to enhance histograms of the quad tree thresholding and second-order statistics and the context of a posteriori spatial probability estimation have been considered thresholding of spatial dependencies. F. Local Adaptive Method Local methods do not determine a single value of threshold but adapt the threshold value depending upon the local image characteristics. A threshold that is calculated at each pixel characterizes this class of algorithms. The value of the threshold depends upon some local statistics like range, variance, and surface fitting parameters or their logical combinations. It is typical of locally adaptive methods to have several, adjustable parameters. The threshold T(i, j) will be indicated as a function of the coordinates i, j, otherwise the object or background decisions at each pixel will be indicated by the logical variable B(i, j). G. Automatic Thresholding T. Balaji, K. Prabha 192

4 AN EFFICIENT FEATURE BASED IMAGE SEGMENTATION OF SATELLITE IMAGES USING LOCAL THRESHOLD METHOD It is make to segmentation is more robust, the threshold should be automatically selected by the system. This is knowledge about the objects, the application, the environment should be used to choose the threshold automatically. It has the following characteristics. Intensity characteristics of the objects Sizes of the objects Fractions of an image occupied by the objects Number of different types of objects appearing in an image H. Hysteresis Thresholding If there is no clear valley in the histogram of an image, it means that there are several background pixels that have similar gray level value with object pixels and vice versa. Hysteresis thresholding can be used pixels above the high threshold are classified as object and below the low threshold as background. The pixels between the low and high thresholds are classified as object only if they are adjacent to other object pixels. I. Optimal Thresholding Suppose that an image contains only two principal regions (e.g., object and background). We can minimize the number of misclassified pixels if we have some prior knowledge about the distributions of the gray level values that make up the object and the background. Assume that the distribution of gray-level values in each region follows a Gaussian distribution. A drawback of the optimal thresholding method is: Prior probabilities might not be known. Object and background distributions might not be known. IV. PRELIMINARIES Thresholding techniques for remote sensing image systems have also received much attention. Because of the wide range of quality distortions over a single image, a combination of threshold operators is often used, with each operator sensitive to a different type of distortion. For example, combines four linear threshold operators to form a single threshold. An example of the operator is T = ku + c, where u is the average contrast over previously scanned images, and k and c are optimizing parameters. In the sequences, we use the following notation. The histogram and the probability mass function (PMF) of the image are indicated, h(g) and p(g) respectively. The gray scale of g = 0...G, where G is the maximum luminance value in the image, typically 255 if 8-bit quantization is assumed. If the gray value range is not explicitly indicated as [g min, g max] it will be assumed to extend from 0 to G. The cumulative probability function is defined as g P(g) p(i) i 0 It is assumed that the PMF is estimated from the histogram of the image by normalizing to the number of samples at every gray level. In the context of document processing, the foreground (object) is the set of pixels with luminance values less than T, while the background pixels have luminance value above this threshold. In satellite images the foreground area may consists of darker (more absorbent, denser etc.) regions or conversely of shinier regions, for example that hotter, more reflective, less dense regions. In contexts where the object appears brighter than the background the definitions of the foreground and background will be simply toggled. The foreground and background area probabilities are calculated as: T P f (T) P f p(g) P b(t) P b p(g) g 0 and g T 1 The Shannon entropy parametrically dependent upon the threshold value T for the foreground and background is formulated as: H f (T) g 0 H b(t) T G p f (g)log p f (g) p b(g)log p b(g) g T 1 The sum of these two thresholds are expressed as H(T) H f (T) Hb(T) When the entropy is calculated over the input image distribution p(g) then obviously it does not depend upon the threshold T and hence is expressed simply as H. For various other definitions of the entropy in the context of thresholding, with some abuse of notation, we will use the same symbols of H f (T) and H b (T). A single threshold will not work well when we have uneven illumination due to shadows or due to the direction of illumination. The idea is to partition the image into m x m sub images and then choose a threshold Tij for each sub image. This approach might lead to sub images having simpler histogram (e.g. remote sensing image). Fig. 1. Each Sub Images with Simple Histogram Thresholding In case of uneven illumination of histogram, another useful technique is to approximate the values of the image by a simple function called plane. Thersholding can be done relative to the plane (e.g., points above the plane G T. Balaji, K. Prabha 193

5 International Journal of Computer Engineering and Applications, Volume XII, Issue I, Jan. 18, ISSN will be part of the object and anything below will be part of the background). V. PROPOSED LOCAL THRESHOLDING METHOD FOR SATELLITE IMAGE A threshold that is calculated at each pixel characterizes this class of algorithms. The value of the threshold depends upon some local statistics like range, variance, and surface fitting parameters or their logical combinations. In local thresholding, the original remote sensing image is partitioned into smaller sub images and a threshold is determined for each of the sub images. This yields a thresholded image with gray level discontinuities at the boundaries of two different sub images. The threshold of a region can be determined by either the point-dependent method or the region-dependent method. A smoothening technique is then applied to eliminate the discontinuities. In this method, the original image is divided into some consecutive matrix of sub images and a threshold is computed for each sub image. However, a threshold is not computed for sub images with unimodal gray level histogram. Thresholds for such sub images are interpolated from neighboring sub images. For a bimodal sub image, the threshold is computed and to get the better result. First the gray level histogram for a sub image is approximated by a sum of two Gaussian distributions and then the threshold is obtained by minimizing the segmentation error with respect to the threshold value. Finally, the entire thresholded remote sensing image is processed by a low-pass filter to eliminate the gray level discontinuities at the boundaries of sub images. Based on the gray level variation within or between object and background, the gray level co-occurrence matrix is divided into quadrants. Let T h be the threshold within the range 0 T h L-1 that partitions the gray level co-occurrence matrix into four quadrants, namely O, B, O 1 and B 1. The quadrant O represents gray level transition within the object while quadrant B represents gray level transition within the background. The gray level transition between the object and the background or across the objects boundary is placed in quadrant O 1 and quadrant B 1. These four regions can be further grouped into two classes, referred to as local quadrant and joint quadrant. Local quadrant is referred to quadrant O and B as the gray level transition that arises within the object or the background of the image. Then quadrant O 1 and B 1 is referred as joint quadrant because the gray level transition occurs between the object and the background of the image. The local entropic threshold is calculated considering only quadrants O and B. The probabilities of object class and background class are defined as T h T h P O = P ij i=0 j=o A. Algorithm VI. EXPERIMENTAL RESULTS L-1 L-1 P B = P ij i= T h +1 j= T h +1 The local transition entropy of the normalized probabilities of the object class and background class are functions of threshold vector (T h,t h) are defined as POi,j = Pi,j / PO PBi,j = Pi,j / PB The second order local threshold of the object and its background class are functions of threshold vector (T h, T h) are defined as T h T h H O (T h) = -1/2 PO ij log 2 PO ij i=0 j=o L-1 L-1 H B (T h) = -1/2 PB ij log 2 PB ij i= T h +1 j= T h +1 The total second order local threshold of the object and its background class is given by H T(T h) = H O (T h) + H B (T h) Finally, find the T M the gray level corresponding to the maximum of H T(T h) over T h gives the optimal threshold for value T M = avg [Max (H T(T h)] It can be seen that there exists small unconnected pixels in the thresholded image. These isolated pixels are removed by performing length filtering based on connected pixel labeling. The result of removing these unconnected pixels can be seen in the final classified image. To ensure that only the section of the remote sensing image containing data is considered during image processing and analysis, a mask image is generated for each image. It is applied to remove any artifacts present outside the region of interest. The proposed local threshold method for remote sensing image algorithm is shown in Fig. 2. The initial step of the algorithm tests whether the image has a bimodal histogram. If it does, then histogram-based thresholding methods, which have been proven to have outstanding results for bimodal histogram of satellite images. Local threshold methods are particularly effective at saving processing time. This algorithm focuses on remote sensing images of multispectral features. It is a local adaptive analysis method, which uses local feature vectors to find the best approach for thresholding a local area. The original image is recursively broken down into sub regions using quadtree decomposition until an appropriate weighted thresholding method can be applied to each sub region. The outline of the local thresholding algorithm is: T. Balaji, K. Prabha 194

6 AN EFFICIENT FEATURE BASED IMAGE SEGMENTATION OF SATELLITE IMAGES USING LOCAL THRESHOLD METHOD Step 1: Step 2: Step 3: Step 4: Step 5: Step 6: Step 7: Step 8: Test whether the Satellite Image has a Bimodal Decompose Satellite Image into Four Equal Size Local Regions Extract Feature Vectors from Each Local Region Apply Proposed Local Threshold Method for Segmentation Process Repeat Steps 2, 3 and 4 until all Regions are Classified or Categories Define Smoothing of the Edges of Each Region Apply Threshold Method to Each Region Finally the Binary form of Satellite Image is Obtained The various experiment carried out in the above remote sense imagery data set algorithm of LTM in MATLAB 7.6. The complete process of local thresholding and the standard are summarized in subsequent fig. 3. (a) (b) (c) B. Flowchart (d) (e) (f) T. Balaji, K. Prabha 191

7 International Journal of Computer Engineering and Applications, Volume XII, Issue I, Jan. 18, ISSN ( g) ( h) (i) ( j ) ( k ) ( l ) ( m) (n) (o ) ( p ) ( q ) ( r) Fig. 3. a. Original I mage, b. RGB in to G ray S cale I mage, c. Histogram of Image, d. Erode of I mage, 3.e. L abeled B lack and W hite I mage, f. Threshold M apping of I mage, g. Segmentation L abeled I mage1, h. Segmentation L abeled I mage2, i. Segmentation L abeled I mage3, j. Segmentation with J et F actor k. Segmentation with A utumn F actor, l. Segmentation with C ool F actor, m. Segmentation with C opper F actor, n. Segmentation with H ot F actor, o. Segmentation with S pring F actor, p. Segmentation with S ummer F actor, q. Segmentation with W inter F actor, r. Threshold M agic for B inary I mage. 150, Fig. 2 Proposed Local Thresholding Method T. Balaji, K. Prabha 192

8 AN EFFICIENT FEATURE BASED IMAGE SEGMENTATION OF SATELLITE IMAGES USING LOCAL THRESHOLD METHOD VII. COCLUSION A number of techniques have previously been proposed for thresholding document and normal gray scale images. However, none of them can provide ideal results for degraded features of satellite images. The proposed approach is a local threshold analysis method, which uses local feature vectors to find the best approach for thresholding a local region. Appropriate weighted values are selected automatically for thresholding specific types of remote sensing image under investigation. The original image is recursively broken down into sub regions using quad-trees until an appropriate weighted thresholding method can be applied to each of the sub region. This algorithm analyses the features extracted from local regions to determine the appropriate weighted value for a threshold method based on mean gradient. We usually try to classify regions by identifying common properties or identifying differences between regions (edges). From these experimental results, we conclude that the proposed local thresholding method provides better performance that is the thresholded images corresponding to our intuition than that obtained by widely used local threshold method and is useful in the satellite image. It is straightforward to extend the method to multi-level thresholding for satellite imagery data set. DISCUSSIONS The thresholded images obtained by various methods reveal valuable information regarding the thresholding techniques. In this paper a new local thresholding structure called the decompose threshold approach is proposed and compared against some existing local thresholding algorithms for normal gray scale images. From observations during the experiments reported in this paper, degraded remote sensing images normally contain the following characteristics: varying contrast, varying stroke quality, many marks or blotches which do not contain any information. Satisfactory thresholding results can be obtained if only a local method is applied to the whole image. The decompose algorithm is demonstrated as effective at improving the result. It uses local feature vectors to analyze and find the best approach to threshold a local region. Instead of employing a single thresholding algorithm, automatic selection of an appropriate algorithm for specific types of sub regions of the remote sensing image is performed. The original image is recursively broken down into sub regions using quad-tree decomposition until a suitable thresholding method can be applied to each sub region. This paper can accurately classify the different regions so that the appropriate weighted value can be applied to the mean-gradient-based threshold method. These results may be interpreted as due to the more powerful energy minimization. The overall execution times of the proposed method were around seconds for the remote sensing data sets and acceptable times in the usual applications to land cover or remote based satellite mapping of segmentation methods. REFERENCES [1] Balaji T., An Automatic Color Feature Vector Classification Based on Clustering Method, IJRSI, Vol. IV, Issue IV, pp , April [2] Balaji T.,and Sumathi M., Effective Statistical Multilevel Thresholding of Remote Sensing Image Classification, Multidisciplinary Research Journal of VVV College, Vol. II, Issue 1, pp , August [3] Balaji T., Robust and Realistic Classification of Massive Gray level Thresholding in Remote Sensing Images, International Journal of Computer Science and Engineering, Vol. 2, Issue 11, pp , Nov [4] Balaji T., and Sumathi M., Relational Features of Remote Sensing Image Segmentation using Effective K-Means Clustering, IJoART, Vol. 8, Issue 8, pp , August [5] Balaji T., and Sumathi M., Remote Sensing Image Segmentation A Perspective Analysis, International Journal of Third Concept, pp , September [6] Weszka J.S., A Survey of Threshold Selection Techniques, Vision Graphics Image Process, pp , [7] A. Y. Wu and A. Rosenfeld, Threshold Selection using Quad Tree, IEEE Trans., PAMI-14, pp , [8] N. Otsu, A Threshold Selection Method from Gray-level Histograms, IEEE Trans. SMC- 29, pp , [9] Yanowitz, S.D., and Bruckstein, A.M., A New Method for Image Segmentation, Computer Vision Graphics and Image Processing, Vol. 46, pp , [10] W.N. Lie, An Efficient Threshold-Evaluation Algorithm for Image Segmentation Based on Spatial Gray-level Cooccurrences, Signal Processing, Vol. 63, pp , [11] T. Pavlidis, Threshold Selection using Second Derivatives of the Gray-scale Image, International Conference on Document Analysis and Recognition, pp , [12] J.D. Yang, Y.S. Chen, W.H. Hsu, Adaptive Thresholding Algorithm and its Hardware Implementation, Pattern Recognition Letters, Vol. 35, pp , [13] X. Zhao, S.H. Ong, Adaptive Local Thresholding with Fuzzy-Validity Guided Spatial Partitioning, International Conference on Pattern Recognition, pp , [14] F.H.Y. Chan, F.K. Lam, H. Zhu, Adaptive Thresholding by Variational Method, IEEE Trans. Image Processing, Vol. 17, pp , [15] S.D. Yanowitz, A.M. Bruckstein, A New Method for Image Segmentation, Computer Graphics and Image Processing, Vol. 66, pp , T. Balaji, K. Prabha 193