Saliency Detection using Region-Based Incremental Center-Surround Distance

Size: px
Start display at page:

Download "Saliency Detection using Region-Based Incremental Center-Surround Distance"

Transcription

1 Saliency Detection using Region-Based Incremental Center-Surround Distance Minwoo Park Corporate Research and Engineering Eastman Kodak Company Rochester, NY, USA Mrityunjay Kumar Corporate Research and Engineering Eastman Kodak Company Rochester, NY, USA Alexander C. Loui Corporate Research and Engineering Eastman Kodak Company Rochester, NY, USA Abstract A new method to detect salient region(s) in images is proposed in this paper. The proposed approach, which is inspired by object-based visual attention theory, segments the input image into coherent regions and measures region-based center-surround distance (RBCSD) using segmented regions and color histograms. Furthermore, segmented regions are merged such that the RBCSD of the merged region is greater than the individual RBCSD of the component regions through region-based incremental distance (RBCSD+I) process. Due to this RBCSD+I process, merged regions may contain incoherent color regions, which improves the robustness of the proposed approach. The key advantages of the proposed algorithm are: (1) it provides a salient region with plausible object boundaries, (2) it is robust to color incoherency present in the salient region, and (3) it is computationally efficient. Extensive qualitative and quantitative evaluation of the proposed algorithm on widely used data sets and comparison with the existing saliency detection approaches clearly indicates the feasibility of the proposed approach. Keywords-Saliency, center-surround distance, image segmentation. I. INTRODUCTION Visual saliency refers to a distinct subjective perceptual quality that makes certain locations or objects of an image stand out from the surroundings. It is well known that the primate visual system processes only the most important or salient information present in the scene using a selective attention mechanism where salient information usually corresponds to location or objects that exhibit high visual saliency [1], [2], [3]. Saliency detection also works on the similar principle and attempts to model salient objects present in input images or videos. Saliency detection is an important research topic and one can find its application in many image processing and computer vision areas such as object detection [4], object tracking [5], video compression [6], and photorealistic animation [7], to name a few. There are two governing theories in visual attention [3], [8]: (i) space-based approach, and (ii) object-based approach. Space-based approach, a time-honored model for saliency detection, attempts to identify visually salient spatial locations. There are a number of computational models that support space-based theory in psychophysics and computer vision literature [1], [2], [9], [10]. The major difference between these models lies in the methods of computing and (a) (b) (c) (d) (e) (f) (g) (h) Figure 1. (a) Input image (b) Segmentation using [16] (c) Saliency map [13] (d) Saliency object detection using [13] (e) Salient object: ground truth (f) Segmentation result: proposed approach (g) Saliency map: proposed approach (h) Salient object detection: proposed approach combining features to generate continuous saliency maps [3]. The goal of the object-based approach is to select salient objects [8] and numerous algorithms that support this approach have been introduced in recent years in psychophysics and computer vision literature [3], [11], [12], [13], [14], [15]. However, saliency computation of these methods is performed independently from a segmentation and a further segmentation (grouping) is performed based on the saliency map computed by the traditional space-based computational model. Our proposed method is also an object-based computational model. However, one fundamental difference from the aforementioned object-based models is that our method performs segmentation (or perceptual grouping) and saliency computation simultaneously in such a way that the saliency represented by the region-based center-surround distance (RBCSD) 1 of the merged regions is higher than that of each corresponding component region. Therefore, our proposed method enables grouping of inhomogeneous regions as long as the merging of the inhomogeneous regions increases the 1 The RBCSD will be explained in detail in Section III.

2 saliency. As can be seen in Figure 1, the head and face of the chicken are grouped as one object even if they have incoherent colors. Extensive qualitative and quantitative evaluations clearly indicate that our proposed method has the state-of-theart performance in terms of the speed and accuracy over the existing methods when evaluated on a public data set introduced by [13]. II. RELATED WORK Prior methods for identifying salient objects in a digital image generally require a computationally intensive search process. Such methods also typically require each salient object to be composed of homogeneous regions. Itt and Koch [1][2] introduced a biologically motivated model for saliency detection. The key steps of their model include (i) computation of low-level features of images (e.g., color contrast, edges, and edge orientations at different scales using up and down sampling), (ii) computation of center surround responses at different scales using differences of Gaussians, and record local maxima responses, and (iii) combining of all the computed responses to generate a saliency map. However, the saliency map does not provide concrete boundaries around salient regions as this algorithm is based on the space-based visual saliency theory. In [9], Hou et al. described a space-based visual saliency algorithm to approximate the innovation part of an image by removing statistically redundant components. This approach involves a high pass filtering operation of log spectral magnitudes and tends to detect small salient objects and edges of a large salient object. The algorithm proposed by Luo et al. in [11] automatically determines main subjects in photographic images. This method provides a measure of belief for the location of main subjects within a digital image and thereby provides an estimate of the relative importance of different subjects in an image. The output of the algorithm is in the form of a list of segmented regions ranked in descending order of their estimated importance. This algorithm first segments an input image and then groups regions into larger segments that correspond to physically coherent objects. A saliency score is then computed for each of the resulting regions and the region that mostly contains the main subject is determined using probabilistic reasoning. However, one of the shortcomings of this approach is that image regions that constitute the main subject must be coherent. For example, if the main subject is a person wearing a red shirt with black pants, the region merging technique in this method will not combine the two regions as it relies on the segmentation algorithm. Liu et al. in [12] proposed a conditional random field learning based framework to combine a set of local, regional, and global features including multiscale contrast, centersurround histogram, and color spatial distribution to detect salient objects. However, a strong limitation of this method is that it requires a range of values for ratio and rectangular sizes to evaluate the center-surround histogram. Luo et al. [17] addresses the limitation of [12] by searching a rectangular region that maximizes the saliency density. However, [17] can neither define the salient object boundary nor compute the salient rectangular boundary simultaneously. Valenti et al. exploited the fact that, to infer visual saliency, salient objects should have local characteristics different from the rest of the scene [15]. They used a novel operator to combine local characteristics such as edges, color, shape, etc. to deduce global information. The resulting global information is used in conjunction with a segmentation algorithm to discriminate salient objects from the background. The algorithm proposed by Achanta et al. [13] produces a full-resolution saliency map with well-defined boundaries for the salient objects. This algorithm first computes a mean color for the entire image and then the mean color is subtracted from each pixel value to produce the saliency map. Then mean-shift segmentation is performed independently to produce the salient region boundary by discarding image segments that have mean saliency values of the segments lower than an adaptive threshold. This approach is not capable of detecting the local salient region if the mean color of the local salient region is similar to that of the entire image. Also it does not have real time performance. Thus, there remains a need for a computationally efficient and robust method to determine object saliency in a digital image that can work with both homogeneous and nonhomogeneous image regions with well-defined salient object boundaries. Our proposed method addresses these problems effectively and determines salient regions with plausible object boundaries. Furthermore, it is robust to the color incoherency of the salient region and is computationally efficient. Figure 2. III. PROPOSED METHOD Overview of the proposed method. An overview of the proposed approach is shown in Figure 2. The Downsample step first downsamples the input image and the Initial Segmentation step segments the downsampled image into coherent regions. Although we use the segmentation method proposed by [16] due to its speed and simplicity, any off-the-shelf segmentation algorithm can be used at this stage to segment downsampled input image. The Compute RBCSD step computes RBCSD for each

3 segmented region produced by the Initial Segmentation step. We call each segmented region as a center region and the rest of the image region as the surrounding region of the corresponding center region, as illustrated in Figure 3(a) 3(d), where c i is the i th center region and s i is the corresponding surround region. Note that each surrounding region is defined with respect to a particular center region; therefore, no two surrounding regions are the same, i.e., s i s j if c i c j. The dissimilarity between the center region c i and the surrounding region s i is denoted as the RBCSD w i and is computed using distance metric based on three dimensional Hue-Saturation-Value (HSV) color histogram. (a) (c) Figure 3. (a) Center region (c i ) (b) s i : surrounding region corresponding to c i (c) center region (c ij ) obtained by merging regions c i and c j (d) s ij : surrounding region corresponding to c ij. Next, Construct Graph step constructs a graph (G = {C, E}) where center region c i is the i th node of the graph and the weight of the edge connecting nodes c i, and c j is the RBCSD of the region obtained by merging c i, and c j. The RBCSD+I Process step applies the proposed RBCSD+I algorithm to the center regions based graph (generated at step Construct Graph ) to produce the saliency map. In particular, RBCSD+I merges neighboring center regions c i and c j if the merge results in a higher RBCSD as compared to the individual RBCSDs for those regions. After merging, both center regions and the surround regions are updated (see Figure 3c 3d). This process is repeated until no center region is available for merging that will increase the RBCSD. Due to this selective merging, RBCSD+I produces a segment map wherein each segment has a saliency value. Therefore, the segment map directly lead to the image segmentation (see Figure 1(f)) and the saliency values directly lead to the saliency map (see Figure (b) (d) 1(g)). Note that RBCSD+I is a complex process and must be implemented efficiently to achieve good accuracy and real-time performance. Note that the saliency map produced by the RBCSD+I Process step is a gray scale image. Therefore, to extract a salient object from the input image, we binarize the saliency map ( Thresholding step) using an adaptive thresholding [13] (see Figure 1(h)). Detailed descriptions of the image segmentation algorithm, computation of RBCSD, graph construction, and RBCSD+I process are provided next. A. Graph-based Segmentation Algorithm As mentioned previously, input image segmentation is the first step of the proposed approach. We use the graph-based segmentation algorithm (GSA) proposed in [16] to perform image segmentation. In addition to image segmentation, as explained later in this paper, we also use a general graph representation and disjoint-set forests data structure used widely for segmentation algorithms to implement our proposed algorithm. Since GSA also uses a graph representation as well as disjoint-set forests data structure and is used as an initialization step to our proposed method, we provide a brief review of GSA for the completeness of the paper and better understanding. Note that other segmentation algorithms can also easily be used in the proposed framework and we selected GSA solely because of its simplicity and computation efficiency. A sample different behavior of our method and GSA in terms of segmentation can be seen in Figures 1(b) and 1(f). In GSA [16], initially, a graph is constructed over the entire image and each pixel p i is treated as a unique segmented region c i and initial edge is defined over eight connected neighborhoods of each pixel p i. The weight of the edge is the Euclidean distance between the color of the edge pixels. Then each edge is sorted in ascending order by edge weight. By iteratively checking each edge in the sorted order, the merging of two nodes c i and c j connected by the edge is performed when a minimum edge weight that connects c i and c j is smaller than minimal internal variation of c i and c j given as MInt(c i, c j ) = min(int(c i ) + k/ c i, Int(c j ) + k/ c j where Int(c i ) is a maximum weight in c i and c i is the number of pixels in c i. The term k/ c i is used to enable initial merging even if the weight of the edge is larger than their internal variations, since the internal variation of an initial node (pixel p) is zero. The parameter k indirectly controls the size of each segment with a larger k, which generally results in larger segments with a higher probability of missing segmentation boundaries. Interested readers are referred to [16] for more details. We first downsample the input image in such a way that max(h, w) = 64., while we preserve the aspect ratio where h and w are the height and width, respectively, of the downsampled image. We segment the downsampled image

4 into a set of N regions where the i th region is represented as c i (1 i N) and the value of k is set to 10. Figure 4. Graph structure with a node for each center region c i and an edge e ij defined between two neighboring center regions c i and c j The weight of the edge between regions c i and c j, w ij (= w ij ) is computed by the center-surround distance of merged region c ij. B. RBCSD Computation We then compute unnormalized three-dimensional HSV color histograms for the entire image and each center region c i (1 i N), which are denoted as A(I) and A(c i ), respectively. Then the HSV color histogram of surrounding region of c i, A(s i ) can be computed efficiently as: A(s i ) = A(I) A(c i ) (1) Similarly, the HSV color histogram for merged regions c i and c j can be computed efficiently as: A(c ij ) = A(c i ) + A(c j ) (2) and the HSV color histogram of the surrounding region of c ij can be computed as: A(s ij ) = A(I) A(c ij ) (3) Next we compute a center-surround distance w i using a normalized cross correlation between the center region c i and the surrounding region s i, which is formally given as: w i = (1 NCC (A(c i ), A(s i )))/ 2 (4) where N CC(A, B) is a normalized cross correlation function between A and B. This formulation enables fast computation that does not require color histogram computation repeatedly during our proposed RBCSD+I process and it is valid for any histogram-based features without loss of generality. Moreover, once A(I) and A(c i ) for 1 i N are computed, any of the actual merging for c i does not need to be performed during the proposed RBCSD+I process since normalized cross-correlation and the HSV color histogram of any merged region and surrounding region can be computed by linear operations on histograms. We use 10 bins for each of H, S, and V channels resulting in 1000 by one color histogram vector. C. Graph Construction Then we build an undirected graph G = (C, E) where c i C is the center regions and e ij E is an edge between neighboring region c i and c j. An example of such a graph, based on Figure 3(a), is shown in Figure 4. For every edge e ij between neighboring center regions c i and c j, its weight w ij is computed as: w ij = [1 NCC (A (c ij ), A (s ij ))] /2 (5) where A (c ij ) and A (s ij ) are computed by equations (2) and (3). Note that weight w ij in the proposed method considers not only the two neighboring regions but also the surrounding region of the two regions while most of the graph-based segmentation algorithm measures the similarity of the neighboring two regions only. Proposed Algorithm 1 Downsample input image //see Figure 2 Downsample step 2 Segment the downsampled image to get c i for 1 i N 3 Set i = 1, m = 1 4 while i N 5 Set p(i) = i // Set parent node of i as i. 6 Compute w i //see Figure 2 Compute RBCSD step 7 i=i+1 8 end 9 Build a graph G = (C, E), c i C, e m E, e m = e ij = (c i, c j, w ij), (1 m K) //see Figure 2 Construct Graph step 10 Sort e m(= e ij) in the order of decreasing w ij 11 while m K // for e m = e ij = (c i, c j, w ij). 12 a=p(i), b=p(j) // find parent node of i and j 13 Compute w ab 14 if w ab > w a and w ab > w b or NCC(c a, c b ) > Merge c a and c b //no actual merging is performed. 16 A(c a) = A(c a) + A(c b ), A(c b ) = A(c a) 17 p(b) = a, w a = w ab, w b = w ab 18 end 19 m=m+1 20 end //see Figure 2 RBCSD+I process step 21 Build saliency map using c p(i) and w p(i) 22 Run adaptive thresholding to binarize the saliency map //see Figure 2 Thresholding step Figure 5. Pseudo code for our proposed algorithm. To implement this algorithm, disjoint-set forests data structure is useful to maintain information about merged region (line 15). D. RBCSD+I Process For the graph G, we sort edge e ij with respect to its weight w ij and start to examine the edge with the highest w ij. The merge of two neighboring regions c i and c j is performed only when the following criterion given by: w ij > w i and w ij > w j or NCC(c i, c j ) > 0.6 (6)

5 is satisfied. The inequality NCC(c i, c j ) > 0.6 enables merging of two similar regions that were not merged by the segmentation algorithm. In Section IV-D, we also report quantitative results by using only the last inequality (RBCSD only) for merge operation to show the effectiveness of the RBCSD+I process. Finally, the entire algorithm is summarized in Figure 5, where RBCSD+I is implemented using the disjoint-set forests data structure. The last line in Figure 5 produces the saliency map using the proposed method. Although the proposed algorithm is a greedy algorithm with respect to merging regions to increase RBCSD, the qualitative and quantitative evaluation of the proposed algorithm on widely used data sets and comparison with the existing saliency detection approaches clearly indicates superior performance of the proposed approach. IV. EXPERIMENT We compare six algorithms on a data set introduced by [13] where the ground truth salient region with a detailed boundary is provided for each of the 1000 images in the database. We denote methods of Itti and Coch [1], Ma and Zhang [18], Harel et al. [19], Hou and Zhang [9], Achanta et al. [14], and Achanta et al. [13] as IT, MZ, GB, SR, AC, and IG, respectively, as used in [13]. B. Salient Region Segmentation by Automatic Thresholding and Image Segmentation We also follow the same comparison method as [13]. Because our proposed method simultaneously segments and detects the salient region, we do not perform segmentation as post-processing as required by all other comparable methods. We simply collect regions of which the saliency value is above the automatic threshold T a to detect the most salient region. The automatic threshold T a is given as: T a = ( 1 N N i=1 w i + 4 max 1 i N w i ) 1 5 The first term in equation (7) is the average RBCSD ( 1 N N i=1 w i) and the second term is the maximum RBCSD (max 1 i N w i ). We set T a given by equation 7 after we tested for 50 images to find out the optimal automatic threshold. However, we found that performance of the proposed method is not sensitive to any value T a between the average RBCSD and the maximum RBCSD. The automatic threshold for other comparable methods is the same as in [13]. As can be seen the quantitative results in Figure 7, our proposed method outperforms all 6 comparable methods in terms of R precision, R recall, and F-measure (= (1+α) R recall R precision α R recall +R precision ) except that ours and IG [13] have about the same R recall. We set α = 0.3 as in [13]. (7) A. Salient Region Segmentation by Fixed Thresholding To obtain the salient region segmentation, a fixed threshold T is applied on the salieny map to generate a binary mask. As can be seen in Figures 1(g) and 9, our saliency map has discrete value since each center region has an uniform saliency (RBCSD) value. We first normalize the saliency map of each algorithm to have saliency values in the range [0, 255], then we vary a threshold T to obtain a binary mask for the salient region. The resulting precision versus recall curve is shown in Figure 6 where the precision rate (R precision ) and recall rate (R recall ) are the ratios of correctly detected salient region (TP) to the detected, and TP to the ground truth salient region, respectively. This curve provides a performance characteristic of each algorithm to detect the salient region. All methods achieve R recall = 100% at T = 0 since entire image is selected as a salient region, resulting the lowest R precision. We use the saliency maps provided by [13] for IT, IG, AC, GB, MZ, and SR. As can be seen in Figure 6, when our proposed method has the minimum R recall of 53%, R precision is 89% and R precision of the state-of-the-art algorithm [13] is 78% when R recall is about the same. Again our minimum R recall of 53% is achieved because each region has uniform RBCSD value. However, the minimum R recall of all other methods is near 0% due to continuous saliency maps. Figure 7. Performance of saliency detection when the mean-shift segmentation is applied to each of the 6 algorithms. The mean-shift clustering is not performed for our proposed method, since the result of our method is direct segmentation of the saliency region. The actual R precision, R recall, and F-measure were provided by the Achanta et al.[13] C. Speed Comparison The typical speed of IG [13] for 300 by 400 images is 3.4 seconds in a modern desktop machine (Intel Dual Core 2.4Ghz, 4GB RAM, unoptimized C++ implemented in serial processing), whereas our method only takes seconds under the same machine while exhibiting better performance characteristic in terms of the precision and recall curve than the state-of-the-art algorithm, IG [13] (see Figure 6).

6 Figure 6. Left: Precision and recall curve with respect to varying threshold T. Note that the recall rate of our method is 53% when the best precision rate is 89% for 1000 images. Overall, GB and AC perform similarly. The discontinuity in the curve for our method exists because the threshold T is varied from 0 to 254 and the most salient region in our saliency map has value of 255 (See our saliency maps in Figure 9). Right: Best performance of each algorithm when it is optimized for the best F-measure with respect to varying threshold T. (a) Input, GT, Ours 1, Ours 2 (b) Input, GT, Ours 1, Ours 2 Figure 10. From the left to the right in each sub figure: input image, ground truth (GT), our result for the most salient region, and our result for multiple salient parts with their saliency ranks are shown, respectively (the more salient region has the lower rank). D. Effect of RBCSD+I To show the effectiveness of the RBCSD+I, we also run the proposed algorithm using only the last test clause in the merging criterion given by equation (6). As can be seen in Figure 8, the method using only the last clause generates a different segmentation, resulting different RBCSD value on each region. Quantitative results for both methods for the same test set are as follows: Rprecision of 81% and 83%, Rrecall of 61% and 75%, and F-meaure of 75% and 79% for RBCSD and RBCSD+I, respectively. V. D ISCUSSION As can be seen in some of the results in Figures 9 and 10, the boundaries of salient regions detected by our proposed method are often coarse. Studies on the biological vision system tell us that more refined vision processing occurs selectively when the image is projected onto the retina. Therefore, we seek to improve boundaries of the salient region as a post processing function. In addition, several different salient regions with numeric numbers can be seen in the rightmost column in Figure 10. The lower number indicates the more salient region. The results reveal the semantically hierarchical structure of objects. For example, as can be seen in the rightmost column in Figure 10a, the most salient region is the post box that contains the second most salient region, the white rectangular label and the post box is supported by the third most salient region, black post. As can be seen in the rightmost column in Figure 10b, the arrow within the diamond is ranked as the most salient region and the diamond is ranked as the second most salient region. This shows our proposed method could be used as a visual feature detector for object detection, video key frame detection, and video summarization. We will explore these areas in the future. VI. C ONCLUSION A method to identify high saliency regions in an image is introduced. Our method first segments the downsampled image then measures the region-based center-surround distance using the segmented region and a color histogram. Next, the merge of each segmented region is performed if the centersurround distance of the merged region is higher than those of each other region through RBCSD+I process. Therefore, the regions are merged and center-surround distance of each resulting region is increased even if the merged region might not have coherent color. The benefits of the proposed method are 1) the method is real-time capable, 2) the output is a salient region with plausible object boundaries, and 3) the method tolerates color incoherency of the salient region. The extensive experiments, both qualitative and quantitative, show that our proposed method has the state-of-the-art

7 (a) Input (b) Segmentation by (c) Segmentation by RBCSD RBCSD+I (d) Ground truth (e) Salient region by RBCSD (f) Salient region by RBCSD+I (g) Input (h) Segmentation by RBCSD (i) Segmentation by RBCSD+I (j) Ground truth (k) Salient region by RBCSD (l) Salient region by RBCSD+I Figure 8. Left column: Input image. Second column: Segmentation by RBCSD. Third column: Segmentation by RBCSD+I. Fourth column: Ground truth salient object. Fifth Column: Salient object by RBCSD. Rightmost column: Salient object by RBCSD+I. performance in terms of speed and accuracy over the existing methods when evaluated on a public data set. ACKNOWLEDGMENT The authors thank R. Achanta and co-workers for providing their experimental results. REFERENCES [1] L. Itti, C. Koch, and E. Niebur, A model of saliency-based visual attention for rapid scene analysis, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 2, pp , [2] L. Itti and C. Koch, Computational modelling of visual attention, Nature Reviews. Neuroscience, vol. 2, no. 3, pp , Mar [3] Y. Sun and R. Fisher, Object-based visual attention for computer vision, Artificial Intelligence, vol. 146, pp , May [4] P. Khuwuthyakorn, A. Robles-Kelly, and J. Zhou, Object of interest detection by saliency learning, in 11th European Conference on Computer Vision 2010, pp [5] V. Mahadevan and N. Vasconcelos, Saliency-based discriminant tracking, in IEEE Conference on Computer Vision and Pattern Recognition 2009, pp [6] N. Dhavale and L. Itti, Saliency-based multi-foveated mpeg compression, in Proc. IEEE Seventh International Symposium on Signal Processing and its Applications, Paris, France, Jul 2003, mod;bu;cv, pp [7] L. Itti, N. Dhavale, and F. Pighin, Photorealistic attentionbased gaze animation, in Proc. IEEE International Conference on Multimedia and Expo, Jul 2006, pp [8] G. D. Logan, The code theory of visual attention: an integration of space-based and object-based attention, Psychological Review, vol. 103, no. 4, [9] X. Hou and L. Zhang, Saliency detection: A spectral residual approach, in IEEE Conference on Computer Vision and Pattern Recognition 2007, pp [10] C. Koch and S. Ullman, Shifts in selective visual attention: towards the underlying neural circuitry, Human Neurobiology, vol. 4, no. 4, pp , [11] J. Luo, A. Singhal, S. P. Etz, and R. T. Gray, A computational approach to determination of main subject regions in photographic images, Image and Vision Computing, vol. 22, no. 3, pp , [12] T. Liu, J. Sun, N. ning Zheng, X. Tang, and H. yeung Shum, Learning to detect a salient object, in IEEE Conference on Computer and Vision Pattern Recognition 2007, pp [13] R. Achanta, S. Hemami, F. Estrada, and S. Susstrunk, Frequency-tuned salient region detection, in IEEE Conference on Computer Vision and Pattern Recognition 2009, Jun., pp [14] R. Achanta, F. Estrada, P. Wils, and S. Ssstrunk, Salient region detection and degmentation, in International Conference on Computer Vision Systems 2008, pp [15] R. Valenti, N. Sebe, and T. Gevers, Image saliency by isocentric curvedness and color, in IEEE International Conference on Computer Vision 2009, pp [16] P. F. Felzenszwalb and D. P. Huttenlocher, Efficient graphbased image segmentation, International Journal of Computer Vision, vol. 59, no. 2, pp , [17] Y. Luo, J. Yuan, P. Xue, and Q. Tian, Saliency density maximization for object detection and localization, in Proceedings of the 10th Asian conference on Computer vision Volume Part III, Berlin, Heidelberg, pp [18] Y.-F. Ma and H.-J. Zhang, Contrast-based image attention analysis by using fuzzy growing, in Proceedings of the Eleventh ACM International Conference on Multimedia ACM, 2003, pp [19] J. Harel, C. Koch, and P. Perona, Graph-based visual saliency, in Advances in Neural Information Processing Systems 19, 2007, pp

8 (a) Input, GT, IG [13], Our saliency map, Ours (b) Input, GT, IG [13], Our saliency map, Ours (c) Input, GT, IG [13], Our saliency map, Ours (d) Input, GT, IG [13], Our saliency map, Ours (e) Input, GT, IG [13], Our saliency map, Ours (f) Input, GT, IG [13], Our saliency map, Ours (g) Input, GT, IG [13], Our saliency map, Ours (i) Input, GT, IG [13], Our saliency map, Ours (k) Input, GT, IG [13], Our saliency map, Ours (m) Input, GT, IG [13], Our saliency map, Ours (o) Input, GT, IG [13], Our saliency map, Ours (q) Input, GT, IG [13], Our saliency map, Ours (h) Input, GT, IG [13], Our saliency map, Ours (j) Input, GT, IG [13], Our saliency map, Ours (l) Input, GT, IG [13], Our saliency map, Ours (n) Input, GT, IG [13], Our saliency map, Ours (p) Input, GT, IG [13], Our saliency map, Ours (r) Input, GT, IG [13], Our saliency map, Ours Figure 9. Qualitative comparison to the state-of-the-art algorithm, IG [13]: From left to the right in each sub figure: input image, ground truth (GT), a result of IG[13], our saliency map, and our result are shown, respectively. As can be seen in the figure, our approach detects small and large regions reliably.

Salient Region Detection and Segmentation in Images using Dynamic Mode Decomposition

Salient Region Detection and Segmentation in Images using Dynamic Mode Decomposition Salient Region Detection and Segmentation in Images using Dynamic Mode Decomposition Sikha O K 1, Sachin Kumar S 2, K P Soman 2 1 Department of Computer Science 2 Centre for Computational Engineering and

More information

Main Subject Detection via Adaptive Feature Selection

Main Subject Detection via Adaptive Feature Selection Main Subject Detection via Adaptive Feature Selection Cuong Vu and Damon Chandler Image Coding and Analysis Lab Oklahoma State University Main Subject Detection is easy for human 2 Outline Introduction

More information

Image Compression and Resizing Using Improved Seam Carving for Retinal Images

Image Compression and Resizing Using Improved Seam Carving for Retinal Images Image Compression and Resizing Using Improved Seam Carving for Retinal Images Prabhu Nayak 1, Rajendra Chincholi 2, Dr.Kalpana Vanjerkhede 3 1 PG Student, Department of Electronics and Instrumentation

More information

IMAGE SALIENCY DETECTION VIA MULTI-SCALE STATISTICAL NON-REDUNDANCY MODELING. Christian Scharfenberger, Aanchal Jain, Alexander Wong, and Paul Fieguth

IMAGE SALIENCY DETECTION VIA MULTI-SCALE STATISTICAL NON-REDUNDANCY MODELING. Christian Scharfenberger, Aanchal Jain, Alexander Wong, and Paul Fieguth IMAGE SALIENCY DETECTION VIA MULTI-SCALE STATISTICAL NON-REDUNDANCY MODELING Christian Scharfenberger, Aanchal Jain, Alexander Wong, and Paul Fieguth Department of Systems Design Engineering, University

More information

Robust Frequency-tuned Salient Region Detection

Robust Frequency-tuned Salient Region Detection Robust Frequency-tuned Salient Region Detection 1 Yong Zhang, 2 Yihua Lan, 3 Haozheng Ren, 4 Ming Li 1 School of Computer Engineering, Huaihai Institute of Technology, Lianyungang, China, zhyhglyg@126.com

More information

Image Resizing Based on Gradient Vector Flow Analysis

Image Resizing Based on Gradient Vector Flow Analysis Image Resizing Based on Gradient Vector Flow Analysis Sebastiano Battiato battiato@dmi.unict.it Giovanni Puglisi puglisi@dmi.unict.it Giovanni Maria Farinella gfarinellao@dmi.unict.it Daniele Ravì rav@dmi.unict.it

More information

Saliency Detection for Videos Using 3D FFT Local Spectra

Saliency Detection for Videos Using 3D FFT Local Spectra Saliency Detection for Videos Using 3D FFT Local Spectra Zhiling Long and Ghassan AlRegib School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA ABSTRACT

More information

2.1 Optimized Importance Map

2.1 Optimized Importance Map 3rd International Conference on Multimedia Technology(ICMT 2013) Improved Image Resizing using Seam Carving and scaling Yan Zhang 1, Jonathan Z. Sun, Jingliang Peng Abstract. Seam Carving, the popular

More information

Salient Region Detection and Segmentation

Salient Region Detection and Segmentation Salient Region Detection and Segmentation Radhakrishna Achanta, Francisco Estrada, Patricia Wils, and Sabine Süsstrunk School of Computer and Communication Sciences (I&C), Ecole Polytechnique Fédérale

More information

CS 664 Segmentation. Daniel Huttenlocher

CS 664 Segmentation. Daniel Huttenlocher CS 664 Segmentation Daniel Huttenlocher Grouping Perceptual Organization Structural relationships between tokens Parallelism, symmetry, alignment Similarity of token properties Often strong psychophysical

More information

DETECTION OF IMAGE PAIRS USING CO-SALIENCY MODEL

DETECTION OF IMAGE PAIRS USING CO-SALIENCY MODEL DETECTION OF IMAGE PAIRS USING CO-SALIENCY MODEL N S Sandhya Rani 1, Dr. S. Bhargavi 2 4th sem MTech, Signal Processing, S. J. C. Institute of Technology, Chickballapur, Karnataka, India 1 Professor, Dept

More information

An Improved Image Resizing Approach with Protection of Main Objects

An Improved Image Resizing Approach with Protection of Main Objects An Improved Image Resizing Approach with Protection of Main Objects Chin-Chen Chang National United University, Miaoli 360, Taiwan. *Corresponding Author: Chun-Ju Chen National United University, Miaoli

More information

International Conference on Information Sciences, Machinery, Materials and Energy (ICISMME 2015)

International Conference on Information Sciences, Machinery, Materials and Energy (ICISMME 2015) International Conference on Information Sciences, Machinery, Materials and Energy (ICISMME 2015) Brief Analysis on Typical Image Saliency Detection Methods Wenwen Pan, Xiaofei Sun, Xia Wang, Wei Zhang

More information

Salient Region Detection using Weighted Feature Maps based on the Human Visual Attention Model

Salient Region Detection using Weighted Feature Maps based on the Human Visual Attention Model Salient Region Detection using Weighted Feature Maps based on the Human Visual Attention Model Yiqun Hu 2, Xing Xie 1, Wei-Ying Ma 1, Liang-Tien Chia 2 and Deepu Rajan 2 1 Microsoft Research Asia 5/F Sigma

More information

Small Object Segmentation Based on Visual Saliency in Natural Images

Small Object Segmentation Based on Visual Saliency in Natural Images J Inf Process Syst, Vol.9, No.4, pp.592-601, December 2013 http://dx.doi.org/10.3745/jips.2013.9.4.592 pissn 1976-913X eissn 2092-805X Small Object Segmentation Based on Visual Saliency in Natural Images

More information

A Novel Approach for Saliency Detection based on Multiscale Phase Spectrum

A Novel Approach for Saliency Detection based on Multiscale Phase Spectrum A Novel Approach for Saliency Detection based on Multiscale Phase Spectrum Deepak Singh Department of Electronics & Communication National Institute of Technology Rourkela 769008, Odisha, India Email:

More information

FACULTY OF ENGINEERING AND INFORMATION TECHNOLOGY DEPARTMENT OF COMPUTER SCIENCE. Project Plan

FACULTY OF ENGINEERING AND INFORMATION TECHNOLOGY DEPARTMENT OF COMPUTER SCIENCE. Project Plan FACULTY OF ENGINEERING AND INFORMATION TECHNOLOGY DEPARTMENT OF COMPUTER SCIENCE Project Plan Structured Object Recognition for Content Based Image Retrieval Supervisors: Dr. Antonio Robles Kelly Dr. Jun

More information

A Model of Dynamic Visual Attention for Object Tracking in Natural Image Sequences

A Model of Dynamic Visual Attention for Object Tracking in Natural Image Sequences Published in Computational Methods in Neural Modeling. (In: Lecture Notes in Computer Science) 2686, vol. 1, 702-709, 2003 which should be used for any reference to this work 1 A Model of Dynamic Visual

More information

Frequency-tuned Salient Region Detection

Frequency-tuned Salient Region Detection Frequency-tuned Salient Region Detection Radhakrishna Achanta, Sheila Hemami, Francisco Estrada, and Sabine Süsstrunk School of Computer and Communication Sciences (IC) Ecole Polytechnique Fédérale de

More information

Unsupervised Saliency Estimation based on Robust Hypotheses

Unsupervised Saliency Estimation based on Robust Hypotheses Utah State University DigitalCommons@USU Computer Science Faculty and Staff Publications Computer Science 3-2016 Unsupervised Saliency Estimation based on Robust Hypotheses Fei Xu Utah State University,

More information

PERFORMANCE ANALYSIS OF COMPUTING TECHNIQUES FOR IMAGE DISPARITY IN STEREO IMAGE

PERFORMANCE ANALYSIS OF COMPUTING TECHNIQUES FOR IMAGE DISPARITY IN STEREO IMAGE PERFORMANCE ANALYSIS OF COMPUTING TECHNIQUES FOR IMAGE DISPARITY IN STEREO IMAGE Rakesh Y. Department of Electronics and Communication Engineering, SRKIT, Vijayawada, India E-Mail: rakesh.yemineni@gmail.com

More information

CS 664 Slides #11 Image Segmentation. Prof. Dan Huttenlocher Fall 2003

CS 664 Slides #11 Image Segmentation. Prof. Dan Huttenlocher Fall 2003 CS 664 Slides #11 Image Segmentation Prof. Dan Huttenlocher Fall 2003 Image Segmentation Find regions of image that are coherent Dual of edge detection Regions vs. boundaries Related to clustering problems

More information

Synthetic Saliency. Ben Weems Stanford University. Anthony Perez Stanford University. Karan Rai Stanford University. Abstract. 1.

Synthetic Saliency. Ben Weems Stanford University. Anthony Perez Stanford University. Karan Rai Stanford University. Abstract. 1. Synthetic Saliency Anthony Perez Stanford University aperez8@stanford.edu Karan Rai Stanford University karanrai@stanford.edu Ben Weems Stanford University bweems@stanford.edu Abstract There are many existing

More information

Data-driven Saliency Region Detection Based on Undirected Graph Ranking

Data-driven Saliency Region Detection Based on Undirected Graph Ranking Data-driven Saliency Region Detection Based on Undirected Graph Ranking Wenjie Zhang ; Qingyu Xiong 2 ; Shunhan Chen 3 College of Automation, 2 the School of Software Engineering, 3 College of Information

More information

Saliency Detection in Aerial Imagery

Saliency Detection in Aerial Imagery Saliency Detection in Aerial Imagery using Multi-scale SLIC Segmentation Samir Sahli 1, Daniel A. Lavigne 2 and Yunlong Sheng 1 1- COPL, Image Science group, Laval University, Quebec, Canada 2- Defence

More information

An Efficient Salient Feature Extraction by Using Saliency Map Detection with Modified K-Means Clustering Technique

An Efficient Salient Feature Extraction by Using Saliency Map Detection with Modified K-Means Clustering Technique International Journal of Computational Engineering & Management, Vol. 15 Issue 5, September 2012 www..org 63 An Efficient Salient Feature Extraction by Using Saliency Map Detection with Modified K-Means

More information

STUDYING THE FEASIBILITY AND IMPORTANCE OF GRAPH-BASED IMAGE SEGMENTATION TECHNIQUES

STUDYING THE FEASIBILITY AND IMPORTANCE OF GRAPH-BASED IMAGE SEGMENTATION TECHNIQUES 25-29 JATIT. All rights reserved. STUDYING THE FEASIBILITY AND IMPORTANCE OF GRAPH-BASED IMAGE SEGMENTATION TECHNIQUES DR.S.V.KASMIR RAJA, 2 A.SHAIK ABDUL KHADIR, 3 DR.S.S.RIAZ AHAMED. Dean (Research),

More information

The Vehicle Logo Location System based on saliency model

The Vehicle Logo Location System based on saliency model ISSN 746-7659, England, UK Journal of Information and Computing Science Vol. 0, No. 3, 205, pp. 73-77 The Vehicle Logo Location System based on saliency model Shangbing Gao,2, Liangliang Wang, Hongyang

More information

Time Stamp Detection and Recognition in Video Frames

Time Stamp Detection and Recognition in Video Frames Time Stamp Detection and Recognition in Video Frames Nongluk Covavisaruch and Chetsada Saengpanit Department of Computer Engineering, Chulalongkorn University, Bangkok 10330, Thailand E-mail: nongluk.c@chula.ac.th

More information

Texture Segmentation by Windowed Projection

Texture Segmentation by Windowed Projection Texture Segmentation by Windowed Projection 1, 2 Fan-Chen Tseng, 2 Ching-Chi Hsu, 2 Chiou-Shann Fuh 1 Department of Electronic Engineering National I-Lan Institute of Technology e-mail : fctseng@ccmail.ilantech.edu.tw

More information

Combining Top-down and Bottom-up Segmentation

Combining Top-down and Bottom-up Segmentation Combining Top-down and Bottom-up Segmentation Authors: Eran Borenstein, Eitan Sharon, Shimon Ullman Presenter: Collin McCarthy Introduction Goal Separate object from background Problems Inaccuracies Top-down

More information

A New Framework for Multiscale Saliency Detection Based on Image Patches

A New Framework for Multiscale Saliency Detection Based on Image Patches Neural Process Lett (2013) 38:361 374 DOI 10.1007/s11063-012-9276-3 A New Framework for Multiscale Saliency Detection Based on Image Patches Jingbo Zhou Zhong Jin Published online: 8 January 2013 Springer

More information

A Survey on Detecting Image Visual Saliency

A Survey on Detecting Image Visual Saliency 1/29 A Survey on Detecting Image Visual Saliency Hsin-Ho Yeh Institute of Information Science, Acamedic Sinica, Taiwan {hhyeh}@iis.sinica.edu.tw 2010/12/09 2/29 Outline 1 Conclusions 3/29 What is visual

More information

Graph-Based Superpixel Labeling for Enhancement of Online Video Segmentation

Graph-Based Superpixel Labeling for Enhancement of Online Video Segmentation Graph-Based Superpixel Labeling for Enhancement of Online Video Segmentation Alaa E. Abdel-Hakim Electrical Engineering Department Assiut University Assiut, Egypt alaa.aly@eng.au.edu.eg Mostafa Izz Cairo

More information

Feature extraction. Bi-Histogram Binarization Entropy. What is texture Texture primitives. Filter banks 2D Fourier Transform Wavlet maxima points

Feature extraction. Bi-Histogram Binarization Entropy. What is texture Texture primitives. Filter banks 2D Fourier Transform Wavlet maxima points Feature extraction Bi-Histogram Binarization Entropy What is texture Texture primitives Filter banks 2D Fourier Transform Wavlet maxima points Edge detection Image gradient Mask operators Feature space

More information

International Journal of Mechatronics, Electrical and Computer Technology

International Journal of Mechatronics, Electrical and Computer Technology An Efficient Importance Map for Content Aware Image Resizing Abstract Ahmad Absetan 1* and Mahdi Nooshyar 2 1 Faculty of Engineering, University of MohagheghArdabili, Ardabil, Iran 2 Faculty of Engineering,

More information

Object Extraction Using Image Segmentation and Adaptive Constraint Propagation

Object Extraction Using Image Segmentation and Adaptive Constraint Propagation Object Extraction Using Image Segmentation and Adaptive Constraint Propagation 1 Rajeshwary Patel, 2 Swarndeep Saket 1 Student, 2 Assistant Professor 1 2 Department of Computer Engineering, 1 2 L. J. Institutes

More information

BRACE: A Paradigm For the Discretization of Continuously Valued Data

BRACE: A Paradigm For the Discretization of Continuously Valued Data Proceedings of the Seventh Florida Artificial Intelligence Research Symposium, pp. 7-2, 994 BRACE: A Paradigm For the Discretization of Continuously Valued Data Dan Ventura Tony R. Martinez Computer Science

More information

Supplementary Material for submission 2147: Traditional Saliency Reloaded: A Good Old Model in New Shape

Supplementary Material for submission 2147: Traditional Saliency Reloaded: A Good Old Model in New Shape Supplementary Material for submission 247: Traditional Saliency Reloaded: A Good Old Model in New Shape Simone Frintrop, Thomas Werner, and Germán M. García Institute of Computer Science III Rheinische

More information

Color Image Segmentation

Color Image Segmentation Color Image Segmentation Yining Deng, B. S. Manjunath and Hyundoo Shin* Department of Electrical and Computer Engineering University of California, Santa Barbara, CA 93106-9560 *Samsung Electronics Inc.

More information

Including the Size of Regions in Image Segmentation by Region Based Graph

Including the Size of Regions in Image Segmentation by Region Based Graph International Journal of Emerging Engineering Research and Technology Volume 3, Issue 4, April 2015, PP 81-85 ISSN 2349-4395 (Print) & ISSN 2349-4409 (Online) Including the Size of Regions in Image Segmentation

More information

HIGH RESOLUTION REMOTE SENSING IMAGE SEGMENTATION BASED ON GRAPH THEORY AND FRACTAL NET EVOLUTION APPROACH

HIGH RESOLUTION REMOTE SENSING IMAGE SEGMENTATION BASED ON GRAPH THEORY AND FRACTAL NET EVOLUTION APPROACH HIGH RESOLUTION REMOTE SENSING IMAGE SEGMENTATION BASED ON GRAPH THEORY AND FRACTAL NET EVOLUTION APPROACH Yi Yang, Haitao Li, Yanshun Han, Haiyan Gu Key Laboratory of Geo-informatics of State Bureau of

More information

Evaluation of regions-of-interest based attention algorithms using a probabilistic measure

Evaluation of regions-of-interest based attention algorithms using a probabilistic measure Evaluation of regions-of-interest based attention algorithms using a probabilistic measure Martin Clauss, Pierre Bayerl and Heiko Neumann University of Ulm, Dept. of Neural Information Processing, 89081

More information

Hierarchical Saliency Detection Supplementary Material

Hierarchical Saliency Detection Supplementary Material Hierarchical Saliency Detection Supplementary Material Qiong Yan Li Xu Jianping Shi Jiaya Jia The Chinese University of Hong Kong {qyan,xuli,pshi,leoia}@cse.cuhk.edu.hk http://www.cse.cuhk.edu.hk/leoia/proects/hsaliency/

More information

CHAPTER 6 PERCEPTUAL ORGANIZATION BASED ON TEMPORAL DYNAMICS

CHAPTER 6 PERCEPTUAL ORGANIZATION BASED ON TEMPORAL DYNAMICS CHAPTER 6 PERCEPTUAL ORGANIZATION BASED ON TEMPORAL DYNAMICS This chapter presents a computational model for perceptual organization. A figure-ground segregation network is proposed based on a novel boundary

More information

AUTONOMOUS IMAGE EXTRACTION AND SEGMENTATION OF IMAGE USING UAV S

AUTONOMOUS IMAGE EXTRACTION AND SEGMENTATION OF IMAGE USING UAV S AUTONOMOUS IMAGE EXTRACTION AND SEGMENTATION OF IMAGE USING UAV S Radha Krishna Rambola, Associate Professor, NMIMS University, India Akash Agrawal, Student at NMIMS University, India ABSTRACT Due to the

More information

Improving the Efficiency of Fast Using Semantic Similarity Algorithm

Improving the Efficiency of Fast Using Semantic Similarity Algorithm International Journal of Scientific and Research Publications, Volume 4, Issue 1, January 2014 1 Improving the Efficiency of Fast Using Semantic Similarity Algorithm D.KARTHIKA 1, S. DIVAKAR 2 Final year

More information

Detecting Salient Contours Using Orientation Energy Distribution. Part I: Thresholding Based on. Response Distribution

Detecting Salient Contours Using Orientation Energy Distribution. Part I: Thresholding Based on. Response Distribution Detecting Salient Contours Using Orientation Energy Distribution The Problem: How Does the Visual System Detect Salient Contours? CPSC 636 Slide12, Spring 212 Yoonsuck Choe Co-work with S. Sarma and H.-C.

More information

Video saliency detection by spatio-temporal sampling and sparse matrix decomposition

Video saliency detection by spatio-temporal sampling and sparse matrix decomposition Video saliency detection by spatio-temporal sampling and sparse matrix decomposition * Faculty of Information Science and Engineering Ningbo University Ningbo 315211 CHINA shaofeng@nbu.edu.cn Abstract:

More information

Image Retargetting on Video Based Detection

Image Retargetting on Video Based Detection RESEARCH ARTICLE OPEN Image Retargetting on Video Based Detection ALOK THAKUR, NEERAJ TIWARI Electronics And Communication College-Tit Bhopal Emai-Aloksinghv@Yahoo.Com Assistant Professor, Electronics

More information

CHAPTER 6 QUANTITATIVE PERFORMANCE ANALYSIS OF THE PROPOSED COLOR TEXTURE SEGMENTATION ALGORITHMS

CHAPTER 6 QUANTITATIVE PERFORMANCE ANALYSIS OF THE PROPOSED COLOR TEXTURE SEGMENTATION ALGORITHMS 145 CHAPTER 6 QUANTITATIVE PERFORMANCE ANALYSIS OF THE PROPOSED COLOR TEXTURE SEGMENTATION ALGORITHMS 6.1 INTRODUCTION This chapter analyzes the performance of the three proposed colortexture segmentation

More information

Compression of Stereo Images using a Huffman-Zip Scheme

Compression of Stereo Images using a Huffman-Zip Scheme Compression of Stereo Images using a Huffman-Zip Scheme John Hamann, Vickey Yeh Department of Electrical Engineering, Stanford University Stanford, CA 94304 jhamann@stanford.edu, vickey@stanford.edu Abstract

More information

Dynamic visual attention: competitive versus motion priority scheme

Dynamic visual attention: competitive versus motion priority scheme Dynamic visual attention: competitive versus motion priority scheme Bur A. 1, Wurtz P. 2, Müri R.M. 2 and Hügli H. 1 1 Institute of Microtechnology, University of Neuchâtel, Neuchâtel, Switzerland 2 Perception

More information

Quasi-thematic Features Detection & Tracking. Future Rover Long-Distance Autonomous Navigation

Quasi-thematic Features Detection & Tracking. Future Rover Long-Distance Autonomous Navigation Quasi-thematic Feature Detection And Tracking For Future Rover Long-Distance Autonomous Navigation Authors: Affan Shaukat, Conrad Spiteri, Yang Gao, Said Al-Milli, and Abhinav Bajpai Surrey Space Centre,

More information

Optimal Grouping of Line Segments into Convex Sets 1

Optimal Grouping of Line Segments into Convex Sets 1 Optimal Grouping of Line Segments into Convex Sets 1 B. Parvin and S. Viswanathan Imaging and Distributed Computing Group Information and Computing Sciences Division Lawrence Berkeley National Laboratory,

More information

Short Run length Descriptor for Image Retrieval

Short Run length Descriptor for Image Retrieval CHAPTER -6 Short Run length Descriptor for Image Retrieval 6.1 Introduction In the recent years, growth of multimedia information from various sources has increased many folds. This has created the demand

More information

CS 534: Computer Vision Segmentation and Perceptual Grouping

CS 534: Computer Vision Segmentation and Perceptual Grouping CS 534: Computer Vision Segmentation and Perceptual Grouping Ahmed Elgammal Dept of Computer Science CS 534 Segmentation - 1 Outlines Mid-level vision What is segmentation Perceptual Grouping Segmentation

More information

Histogram and watershed based segmentation of color images

Histogram and watershed based segmentation of color images Histogram and watershed based segmentation of color images O. Lezoray H. Cardot LUSAC EA 2607 IUT Saint-Lô, 120 rue de l'exode, 50000 Saint-Lô, FRANCE Abstract A novel method for color image segmentation

More information

CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION

CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION CHAPTER 6 MODIFIED FUZZY TECHNIQUES BASED IMAGE SEGMENTATION 6.1 INTRODUCTION Fuzzy logic based computational techniques are becoming increasingly important in the medical image analysis arena. The significant

More information

Content Based Image Retrieval: Survey and Comparison between RGB and HSV model

Content Based Image Retrieval: Survey and Comparison between RGB and HSV model Content Based Image Retrieval: Survey and Comparison between RGB and HSV model Simardeep Kaur 1 and Dr. Vijay Kumar Banga 2 AMRITSAR COLLEGE OF ENGG & TECHNOLOGY, Amritsar, India Abstract Content based

More information

A New Feature Local Binary Patterns (FLBP) Method

A New Feature Local Binary Patterns (FLBP) Method A New Feature Local Binary Patterns (FLBP) Method Jiayu Gu and Chengjun Liu The Department of Computer Science, New Jersey Institute of Technology, Newark, NJ 07102, USA Abstract - This paper presents

More information

Object Tracking Algorithm based on Combination of Edge and Color Information

Object Tracking Algorithm based on Combination of Edge and Color Information Object Tracking Algorithm based on Combination of Edge and Color Information 1 Hsiao-Chi Ho ( 賀孝淇 ), 2 Chiou-Shann Fuh ( 傅楸善 ), 3 Feng-Li Lian ( 連豊力 ) 1 Dept. of Electronic Engineering National Taiwan

More information

A Novel Approach to Saliency Detection Model and Its Applications in Image Compression

A Novel Approach to Saliency Detection Model and Its Applications in Image Compression RESEARCH ARTICLE OPEN ACCESS A Novel Approach to Saliency Detection Model and Its Applications in Image Compression Miss. Radhika P. Fuke 1, Mr. N. V. Raut 2 1 Assistant Professor, Sipna s College of Engineering

More information

Use of Shape Deformation to Seamlessly Stitch Historical Document Images

Use of Shape Deformation to Seamlessly Stitch Historical Document Images Use of Shape Deformation to Seamlessly Stitch Historical Document Images Wei Liu Wei Fan Li Chen Jun Sun Satoshi Naoi In China, efforts are being made to preserve historical documents in the form of digital

More information

Segmentation of Images

Segmentation of Images Segmentation of Images SEGMENTATION If an image has been preprocessed appropriately to remove noise and artifacts, segmentation is often the key step in interpreting the image. Image segmentation is a

More information

Saliency Segmentation based on Learning and Graph Cut Refinement

Saliency Segmentation based on Learning and Graph Cut Refinement MEHRANI, VEKSLER: SALIENCY SEGM. BASED ON LEARNING AND GC REFINEMENT 1 Saliency Segmentation based on Learning and Graph Cut Refinement Paria Mehrani pmehrani@uwo.ca Olga Veksler olga@csd.uwo.ca Department

More information

Using the Kolmogorov-Smirnov Test for Image Segmentation

Using the Kolmogorov-Smirnov Test for Image Segmentation Using the Kolmogorov-Smirnov Test for Image Segmentation Yong Jae Lee CS395T Computational Statistics Final Project Report May 6th, 2009 I. INTRODUCTION Image segmentation is a fundamental task in computer

More information

An Efficient Semantic Image Retrieval based on Color and Texture Features and Data Mining Techniques

An Efficient Semantic Image Retrieval based on Color and Texture Features and Data Mining Techniques An Efficient Semantic Image Retrieval based on Color and Texture Features and Data Mining Techniques Doaa M. Alebiary Department of computer Science, Faculty of computers and informatics Benha University

More information

CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS

CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS CHAPTER 4 CLASSIFICATION WITH RADIAL BASIS AND PROBABILISTIC NEURAL NETWORKS 4.1 Introduction Optical character recognition is one of

More information

2 Proposed Methodology

2 Proposed Methodology 3rd International Conference on Multimedia Technology(ICMT 2013) Object Detection in Image with Complex Background Dong Li, Yali Li, Fei He, Shengjin Wang 1 State Key Laboratory of Intelligent Technology

More information

Supplementary Materials for Salient Object Detection: A

Supplementary Materials for Salient Object Detection: A Supplementary Materials for Salient Object Detection: A Discriminative Regional Feature Integration Approach Huaizu Jiang, Zejian Yuan, Ming-Ming Cheng, Yihong Gong Nanning Zheng, and Jingdong Wang Abstract

More information

A Hierarchial Model for Visual Perception

A Hierarchial Model for Visual Perception A Hierarchial Model for Visual Perception Bolei Zhou 1 and Liqing Zhang 2 1 MOE-Microsoft Laboratory for Intelligent Computing and Intelligent Systems, and Department of Biomedical Engineering, Shanghai

More information

FASA: Fast, Accurate, and Size-Aware Salient Object Detection

FASA: Fast, Accurate, and Size-Aware Salient Object Detection FASA: Fast, Accurate, and Size-Aware Salient Object Detection Gökhan Yildirim, Sabine Süsstrunk School of Computer and Communication Sciences École Polytechnique Fédérale de Lausanne Abstract. Fast and

More information

A Keypoint Descriptor Inspired by Retinal Computation

A Keypoint Descriptor Inspired by Retinal Computation A Keypoint Descriptor Inspired by Retinal Computation Bongsoo Suh, Sungjoon Choi, Han Lee Stanford University {bssuh,sungjoonchoi,hanlee}@stanford.edu Abstract. The main goal of our project is to implement

More information

An Introduction to Content Based Image Retrieval

An Introduction to Content Based Image Retrieval CHAPTER -1 An Introduction to Content Based Image Retrieval 1.1 Introduction With the advancement in internet and multimedia technologies, a huge amount of multimedia data in the form of audio, video and

More information

OTCYMIST: Otsu-Canny Minimal Spanning Tree for Born-Digital Images

OTCYMIST: Otsu-Canny Minimal Spanning Tree for Born-Digital Images OTCYMIST: Otsu-Canny Minimal Spanning Tree for Born-Digital Images Deepak Kumar and A G Ramakrishnan Medical Intelligence and Language Engineering Laboratory Department of Electrical Engineering, Indian

More information

Image Retargeting for Small Display Devices

Image Retargeting for Small Display Devices Image Retargeting for Small Display Devices Chanho Jung and Changick Kim Department of Electrical Engineering Korea Advanced Institute of Science and Technology (KAIST) Daejeon, Republic of Korea ABSTRACT

More information

Color Local Texture Features Based Face Recognition

Color Local Texture Features Based Face Recognition Color Local Texture Features Based Face Recognition Priyanka V. Bankar Department of Electronics and Communication Engineering SKN Sinhgad College of Engineering, Korti, Pandharpur, Maharashtra, India

More information

Robust and Efficient Saliency Modeling from Image Co-Occurrence Histograms

Robust and Efficient Saliency Modeling from Image Co-Occurrence Histograms IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 36, NO. 1, JANUARY 2014 195 Robust and Efficient Saliency Modeling from Image Co-Occurrence Histograms Shijian Lu, Cheston Tan, and

More information

Salient Regions Detection for Indoor Robots using RGB-D Data

Salient Regions Detection for Indoor Robots using RGB-D Data 2015 IEEE International Conference on Robotics and Automation (ICRA) Washington State Convention Center Seattle, Washington, May 26-30, 2015 Salient Regions Detection for Indoor Robots using RGB-D Data

More information

Colour Segmentation-based Computation of Dense Optical Flow with Application to Video Object Segmentation

Colour Segmentation-based Computation of Dense Optical Flow with Application to Video Object Segmentation ÖGAI Journal 24/1 11 Colour Segmentation-based Computation of Dense Optical Flow with Application to Video Object Segmentation Michael Bleyer, Margrit Gelautz, Christoph Rhemann Vienna University of Technology

More information

Graph based Image Segmentation using improved SLIC Superpixel algorithm

Graph based Image Segmentation using improved SLIC Superpixel algorithm Graph based Image Segmentation using improved SLIC Superpixel algorithm Prasanna Regmi 1, B.J.M. Ravi Kumar 2 1 Computer Science and Systems Engineering, Andhra University, AP, India 2 Computer Science

More information

Biometrics Technology: Image Processing & Pattern Recognition (by Dr. Dickson Tong)

Biometrics Technology: Image Processing & Pattern Recognition (by Dr. Dickson Tong) Biometrics Technology: Image Processing & Pattern Recognition (by Dr. Dickson Tong) References: [1] http://homepages.inf.ed.ac.uk/rbf/hipr2/index.htm [2] http://www.cs.wisc.edu/~dyer/cs540/notes/vision.html

More information

AN EFFICIENT BATIK IMAGE RETRIEVAL SYSTEM BASED ON COLOR AND TEXTURE FEATURES

AN EFFICIENT BATIK IMAGE RETRIEVAL SYSTEM BASED ON COLOR AND TEXTURE FEATURES AN EFFICIENT BATIK IMAGE RETRIEVAL SYSTEM BASED ON COLOR AND TEXTURE FEATURES 1 RIMA TRI WAHYUNINGRUM, 2 INDAH AGUSTIEN SIRADJUDDIN 1, 2 Department of Informatics Engineering, University of Trunojoyo Madura,

More information

2D image segmentation based on spatial coherence

2D image segmentation based on spatial coherence 2D image segmentation based on spatial coherence Václav Hlaváč Czech Technical University in Prague Center for Machine Perception (bridging groups of the) Czech Institute of Informatics, Robotics and Cybernetics

More information

TEXT DETECTION AND RECOGNITION IN CAMERA BASED IMAGES

TEXT DETECTION AND RECOGNITION IN CAMERA BASED IMAGES TEXT DETECTION AND RECOGNITION IN CAMERA BASED IMAGES Mr. Vishal A Kanjariya*, Mrs. Bhavika N Patel Lecturer, Computer Engineering Department, B & B Institute of Technology, Anand, Gujarat, India. ABSTRACT:

More information

Image Segmentation Techniques for Object-Based Coding

Image Segmentation Techniques for Object-Based Coding Image Techniques for Object-Based Coding Junaid Ahmed, Joseph Bosworth, and Scott T. Acton The Oklahoma Imaging Laboratory School of Electrical and Computer Engineering Oklahoma State University {ajunaid,bosworj,sacton}@okstate.edu

More information

Learning and Inferring Depth from Monocular Images. Jiyan Pan April 1, 2009

Learning and Inferring Depth from Monocular Images. Jiyan Pan April 1, 2009 Learning and Inferring Depth from Monocular Images Jiyan Pan April 1, 2009 Traditional ways of inferring depth Binocular disparity Structure from motion Defocus Given a single monocular image, how to infer

More information

An Efficient Single Chord-based Accumulation Technique (SCA) to Detect More Reliable Corners

An Efficient Single Chord-based Accumulation Technique (SCA) to Detect More Reliable Corners An Efficient Single Chord-based Accumulation Technique (SCA) to Detect More Reliable Corners Mohammad Asiful Hossain, Abdul Kawsar Tushar, and Shofiullah Babor Computer Science and Engineering Department,

More information

Experimentation on the use of Chromaticity Features, Local Binary Pattern and Discrete Cosine Transform in Colour Texture Analysis

Experimentation on the use of Chromaticity Features, Local Binary Pattern and Discrete Cosine Transform in Colour Texture Analysis Experimentation on the use of Chromaticity Features, Local Binary Pattern and Discrete Cosine Transform in Colour Texture Analysis N.Padmapriya, Ovidiu Ghita, and Paul.F.Whelan Vision Systems Laboratory,

More information

An ICA based Approach for Complex Color Scene Text Binarization

An ICA based Approach for Complex Color Scene Text Binarization An ICA based Approach for Complex Color Scene Text Binarization Siddharth Kherada IIIT-Hyderabad, India siddharth.kherada@research.iiit.ac.in Anoop M. Namboodiri IIIT-Hyderabad, India anoop@iiit.ac.in

More information

Wavelet Based Image Retrieval Method

Wavelet Based Image Retrieval Method Wavelet Based Image Retrieval Method Kohei Arai Graduate School of Science and Engineering Saga University Saga City, Japan Cahya Rahmad Electronic Engineering Department The State Polytechnics of Malang,

More information

Improving Image Segmentation Quality Via Graph Theory

Improving Image Segmentation Quality Via Graph Theory International Symposium on Computers & Informatics (ISCI 05) Improving Image Segmentation Quality Via Graph Theory Xiangxiang Li, Songhao Zhu School of Automatic, Nanjing University of Post and Telecommunications,

More information

A Modified Approach to Biologically Motivated Saliency Mapping

A Modified Approach to Biologically Motivated Saliency Mapping A Modified Approach to Biologically Motivated Saliency Mapping Shane Grant Department of Computer Science University of California, San Diego La Jolla, CA 9093 wgrant@ucsd.edu Kevin A Heins Department

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

IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 14, NO. 4, AUGUST

IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 14, NO. 4, AUGUST IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 14, NO. 4, AUGUST 2012 1275 Unsupervised Salient Object Segmentation Based on Kernel Density Estimation and Two-Phase Graph Cut Zhi Liu, Member, IEEE, Ran Shi, Liquan

More information

A Novel Approach to Image Segmentation for Traffic Sign Recognition Jon Jay Hack and Sidd Jagadish

A Novel Approach to Image Segmentation for Traffic Sign Recognition Jon Jay Hack and Sidd Jagadish A Novel Approach to Image Segmentation for Traffic Sign Recognition Jon Jay Hack and Sidd Jagadish Introduction/Motivation: As autonomous vehicles, such as Google s self-driving car, have recently become

More information

IJREAT International Journal of Research in Engineering & Advanced Technology, Volume 1, Issue 5, Oct-Nov, 2013 ISSN:

IJREAT International Journal of Research in Engineering & Advanced Technology, Volume 1, Issue 5, Oct-Nov, 2013 ISSN: Semi Automatic Annotation Exploitation Similarity of Pics in i Personal Photo Albums P. Subashree Kasi Thangam 1 and R. Rosy Angel 2 1 Assistant Professor, Department of Computer Science Engineering College,

More information

Local Image Registration: An Adaptive Filtering Framework

Local Image Registration: An Adaptive Filtering Framework Local Image Registration: An Adaptive Filtering Framework Gulcin Caner a,a.murattekalp a,b, Gaurav Sharma a and Wendi Heinzelman a a Electrical and Computer Engineering Dept.,University of Rochester, Rochester,

More information

HYBRID CENTER-SYMMETRIC LOCAL PATTERN FOR DYNAMIC BACKGROUND SUBTRACTION. Gengjian Xue, Li Song, Jun Sun, Meng Wu

HYBRID CENTER-SYMMETRIC LOCAL PATTERN FOR DYNAMIC BACKGROUND SUBTRACTION. Gengjian Xue, Li Song, Jun Sun, Meng Wu HYBRID CENTER-SYMMETRIC LOCAL PATTERN FOR DYNAMIC BACKGROUND SUBTRACTION Gengjian Xue, Li Song, Jun Sun, Meng Wu Institute of Image Communication and Information Processing, Shanghai Jiao Tong University,

More information