Abstract This paper investigates the using of Gray Level Co-Occurrence Matrix (GLCM) based on supervised texture segmentation. In most texture segmentation methods, the processing algorithm is applied to a window of the original image rather than to the entire image using sliding scheme. To attain a good segmentation accuracy especially in the boundaries, optimal size of window is determined, or windows of variant sizes are used. Both options are very time consuming. Here, a new technique is proposed to build an efficient GLCM based texture segmentation system. This scheme uses a fixed window of variant apertures. This will reduce the computation overhead and recourses that required to compute GLCM, and will improve the segmentation accuracy. Image's windows are multiplied with a matrix of local operators. After that, GLCM is computed and features are extracted and classified and the segmented image is produced. In order to reduce the segmentation time, two similarity metrics are used to classify the texture pixels. Euclidean metric is used to find the distance between the current and previous GLCM. If it is above a predefined threshold, then the computation of GLCM descriptors are required. Gaussian metric is used as a distance measure between two GLCM descriptors. Furthermore, a median filter is applied to the segmented image. Finally, the transition and misclassified regions are refined. The proposed system is parallelized and implemented on a multicore processor.

References

[1] Brodatz P., A photographic Album for Artists and Designers, Dover New York, 1996.

[2] Fowers J., Brown G., Cooke P., and Stitt G., “A Performance and Energy Comparison Of Fpgas, Gpus, and Multicores for Sliding-Window Applications,” in Proceedings of the ACM/SIGDA International Symposium on Field Programmable Gate Arrays, Monterey, pp. 47- 56, 2012.

[3] Giannini M., Merola P., and Allegrini A., “Texture Analysis for Urban Areas Classification in High Resolution Satellite Imagery,” Applied Remote Sensing Journal, vol. 2, no. 2, pp. 65-71, 2012.

[4] Haralick M., Shanmugam K., and Dinstein I., “Textural Features for Image Classification,” IEEE Transactions on Systems, Man, and Cybernetics, SMC-vol. 3, no. 6, pp. 610-621, 1973.

[5] Huang X., Zhang L., and Li P., “A Multiscale Feature Fusion Approach for Classification of Very High Resolution Satellite Imagery Based on Wavelet Transform,” International Journal of Remote Sensing, vol. 29, no. 20, pp. 5923-5941, 2008.

[6] Kanter D., “Inside Nehalem: Intel´s future processor and system,” Threshold : 0 180 200 220 misclassified pixels E% : ≈ 0 0.0305 0.2642 1.2718 a) Segmentation results when the proposed algorithm is wholly applied using both similarity metrics (window size=13×13, aperture size=9×9) for different threshold values with median filtering. E%≈ 0 4.53 5.41 b) Comparison of the proposed algorithm (left) with

[11] (middle) and

[19] (right). 0 100 200 300 # core time (sec) parfor 207 128 110 110 121 125 spmd 207 110 80 78 98 112 OpenMp 64 35 25 22 18 18 1 2 3 4 5 6 0 20 40 60 threshold value E% 0 0.031 0.264 1.272 4.688 time reduction% 0 7 17 33 47 th 1 th 2 th 3 th 4 th 5 Figure 5. Segmentation results with comparison to previous works. 16 The International Arab Journal of Information Technology, Vol. 16, No. 1, January 2019 http://www.realworldtech.com/ page.cfm? ArticleID=RWT040208182719, Last Visited, 2008.

[7] Mahmoud A., Elbialy S., Pradhan B., and Buchroithner M., “Field-Based Land Cover Classification Using Terra SAR-X Texture Analysis,” Advances in Space Research, vol. 48, no. 5, pp. 799-805, 2011.

[8] Mohamed A. and Lu J., “Analysis of GLCM Parameters for Textures Classification on UMD Database Images,” in Proceedings of The 5th International Conference on Advanced Communications and Computation, Brussels, pp. 111-116, 2015.

[9] Ouma Y., Ngigi T., and Tateishi R., “On the Optimization and Selection of Wavelet Texture for Feature Extraction from High-Resolution Satellite Imagery with Application Towards Urban-Tree Delineation,” International Journal of Remote Sensing, vol. 27, no. 1, pp. 73-104, 2006.

[10] Puig D. and García M., “Pixel-Based Texture Classification by Integration of Multiple Texture Feature Evaluation Windows,” in Proceedings of Iberian Conference on Pattern Recognition and Image Analysis, Puerto de Andratx, pp. 793-801, 2003.

[11] Rampun A., Strange H., and Zwiggelaar R., “Texture Segmentation Using Different Orientations of GLCM Features,” in Proceedings of 6th International Conference on Computer Vision/Computer Graphics Collaboration Techniques and Applications, Berlin, 2013.

[12] Schroeter P. and Bigun J., “Hierarchical Image Segmentation by Multi-Dimensional Clustering and Orientation Adaptive Boundary Refinement,” Pattern Recognition, vol. 28, no. 5, pp. 695-709, 1995.

[13] Sharma M. and Singh S., “Evaluation of Texture Methods for Image Analysis,” in Proceedings of the 7th Australian and New Zealand Intelligent Information Systems Conference, Perth, pp. 117- 121, 2001.

[14] Siqueira F., Schwartz W., and Pedrini H., “Multi- Scale Level Co-Occurrence Matrices for Texture Description,” Neurocomputing, vol. 120, pp. 336- 345, 2013.

[15] Stojmenovic M., Montero A., and Nayak A., “Colour and Texture Based Pyramidal Image,” International Conference on Audio Language and Image Processing, Shanghai, pp. 778-786, 2010.

[16] Susomboon R., Raicu D., Furst J., and Johnson T., “A Co-Occurrence Texture Semi-Invariance to Direction, Distance and Patient Size,” in Proceedings of SPIE-The International Society for Optical Engineering, San Diego, 2008.

[17] Tsai Y., “Texture Image Segmentation Using Adaptive Gray Level Co-Occurrence Probabilities,” in Proceedings of the International Conference on Image Processing, Computer Vision, and Pattern Recognition, Las Vegas Nevada, 2008.

[18] Wang H., Feng Y., Sa Y., Lu J., Ding J., Zhang J., and Hu X., “Pattern Recognition and Classification of Two Cancer Cell Lines by Diffraction Imaging at Multiple Pixel Distances,” Pattern Recognition, vol. 61, pp. 234-244, 2017.

[19] Zheng Y. and Chen K., “A Hierarchical Algorithm for Multiphase Texture Image Segmentation,” ISRN Signal Processing, vol. 2012, pp. 1-11, 2012. Shefa Dawwd was born in Mosul- Iraq in 1970. He received the B.Sc degree in electronic and communication Engineering, the M.Sc and the Ph.D degree in computer Engineering in 1991, 2000, and 2006, respectively. He is presently a faculty member (Associate Professor) in the computer engineering department/University of Mosul. His main research interests include image & signal processing and their hardware models, parallel computer architecture, hardware implementation and GPU based systems. He has authored more than 30 research papers and book chapters. He has been an editorial member of several national and international journals.