ISSN: 0975-766X CODEN: IJPTFI Available Online through Research Article www.ijptonline.com A COMPARATIVE STUDY ON IMAGE SEGMENTATION TECHNIQUES Rajesh Kaluri* School of Information Technology and Engineering, VIT University, Vellore, India. Email: rajesh.kaluri@vit.ac.in Received on 10-05-2016 Accepted on 09-06-2016 Abstract: Image segmentation plays a vital role in various areas of the computer industry. It is having a unique notion in the image processing and morphing techniques. This paper gives an overview and comparison on the methods used in the image segmentation along with its applications. It also provides the usage of methods and the role of image segmentation in the research fields of computer vision. Keywords: Segmentation, Morphing, Threshold, Edge-based, Region growing. Introduction: Image Segmentation is a process which is used to divide the input image into a set of non-identical regions with respect to the specific characteristics like color, texture, intensity etc., of an image [1]. Segmentation of an image gives an idea to a person that what kind of regions can be seen from the image. Segmentation process mainly deals with the pixels of an image, how the regions can be divided and how the neighbor pixels [2] can be mapped with the first selected pixel. Image segmentation process consigns a tag to every pixel of the image, images with similar tags shares the definite features. The main purpose of the Image segmentation process is to provide the meaningful information of the entire image by analyzing the pixels or regions that are stored in the image. Image segmentation can be used in the various fields, such as medicine [3], textile industry, educational research, Human-computer interaction, detecting the criminals etc. Image Segmentation Techniques: Image segmentation encompasses several number of methods [4], but majorly three methods are being used such as Edgebased, Threshold-based and Region-based. All these three methods of the image segmentation have its own unique IJPT June-2016 Vol. 8 Issue No.2 12712-12717 Page 12712
features. Image segmentation carries many techniques in order to extract the meaningful information which is present in the images. The various Image techniques [4] are Constrained active contour, Convex energy function, unsupervised segmentation, kernel metric, Adaptive FCM and so on. The Hybrid region growing proceedswith the input as a medical image and then it computesthe relation with the Harris method by taking the seeds [3].The selected seed will be identifiedas a sample window of size w*w and its origin found [5],Seed regions will be grown in the cloud model. In general, the edge function will not be computed at the angle of 45 degrees. By applying the mask along with the filtering can be easily overcome with the issues. Image Segmentation Edge based Clustering Compression - methods based Histogram - based Threshold based Graph partitioning Region - growing Fig: Various methods in Image segmentation. Clustering method: This method is mainly used to segment the data of the images. It is having a superior algorithm called K-means algorithm, which will partition the given image into K clusters. It has a basic algorithm with four steps. Step.1: Randomly choose K clusters centers. Step.2: In order to reduce the distance between the pixel of the image and cluster center, allocate separate pixel in the image to the cluster. Step.3: Moderate all the pixels of the cluster and develop the cluster centers. Step.4: Do step 2 and 3 until no pixels has variation in the clusters. The L-dimensional Euclidean space is defined along with the Probability of False Alarm (PFA) by k as follows [6] [7]. IJPT June-2016 Vol. 8 Issue No.2 12712-12717 Page 12713
Edge- based method: It is one of the most important methods in the image segmentation process [8], which focus on the boundary regions of the image. The input image can be segmented by means of the boundary edges along with the object boundaries. It has some limitations like image with flat boundaries, images without proper edges, images with the text boundaries. Threshold method: Threshold method of the image segmentation process will take the input image(s) and partitions the image [9] [10] by using the threshold value T. Each image pixel will be having more than two values. It also been a part of multi-dimensional fuzzy algorithms along with the rules of fuzzy systems. Threshold method margins to give the effective output because of its T value and the segmented image might be lesser or bigger than the actual one. Histogram based Method: It mainly uses only one pass along the pixels, here the pixels of the image will be computed by the histogram [11]. In order to trace the clusters in the image, the histogram peak and valleys are used. The small clusters can be obtained by having the histogram-seeking method recursively on to the clusters in the image [12]. Using single pass efficiency, the Histogram based method [13] can also be applying to the multiple frames. It is a very useful technique for the video tracking systems [14]. Graph partitioning methods: Graph partitioning methods will work to the image segmentation based on its own attributes like undirected graph, weight, minimum cut, minimum spanning tree etc. Markov Random Fields (MRF)is one among the graph partition methods [15] [16] which works the cliques, marginal probability distributions. MRF s image segmentation consists of supervised and unsupervised segmentations. Fig: A chosen pixel in MRF neighborhood. IJPT June-2016 Vol. 8 Issue No.2 12712-12717 Page 12714
Wrapper based segmentation is one that extremely diminishes the need of an object interest tobe segmented must be homogeneous in some low-level image parameter, such as texture, color, or grayscale [12]. Region-based method: Region-based method aims to give better results in the image segmentation process and which is an evolutionary method in many research fields [17]. It primarily identifies a seed point in the image and then recognizes the nearest neighbor pixel. If the nearest neighbor pixel matches with its pixel ratio, then the seed point penetrates into the next pixel and thus a small region will form. Fig: Region growing example [19]. Region Growing by Mean or Max-Min: Region growing is a popular technique in terms of its methods like Meanor Max Min. It can be observed primarily by investigatingthe assets of each block and integrating them with the neighboring blocks. In Mean technique, the mean values of each block will be identified and based on the mean values, the process will be determining whether the blocks to be merged or not. The technique of Max-Min is to find the difference of the two regions and merging with the adjacent regions only if the difference is acceptable by the seed block [18]. The process will be repeated with all the new seed blocks until the difference is acceptable by the seed block. Image segmentation has a wide variety of applications in various areas such as machine vision, object detection (Face and Iris recognition in HCI), health-care, video surveillance, Traffic control systems. In Harris Corner Detector, the seed IJPT June-2016 Vol. 8 Issue No.2 12712-12717 Page 12715
selection can be done automatically and gives an efficient way in time consuming.the auto correlation function with the point (x, y) and shift is given below [19]. Conclusion: Image segmentation carries a tremendous role in the computer vision like digital image and morphing. Major techniques of Image Segmentation will provide a better result especially in medicine. At the outset, the image segmentation has enormous methods to give a meaningful information as an output and each method is having its own feature. The survey shows that many researchers uses Region growing algorithm in their work because of its seed pixel and neighbor pixels. References: 1. S.-C. Cheng, Region-growing approach to colour segmentationusing 3-D clustering and relaxation labelling, IEE Proc.-Vis. Image Signal Process., Vol. 150, No. 4, August 2003. 2. J. Alison Noble, and Djamal Boukerroui, Ultrasound Image Segmentation: A Survey, IEEE Transactions on Medical Imaging, vol. 25, no. 8, August 2006. 3. D. Muhammad Noorul Mubarak et[al], A Hybrid Region Growing algorithm for Medical image segmentation, International Journal of Computer Science & Information Technology (IJCSIT) Vol 4, No 3, June 2012. 4. H. Narkhede, Review of image segmentation techniques, Int. J.Sci. Mod. Eng, vol. 1, p. 28, 2013. 5. Arti Taneja, PriyaRanjan, Amit Ujjlayan, A Performance Study of Image Segmentation Techniques in IEEE Conference, 2015. 6. J. F. Khan, S. M. A. Bhuiyan, and R. R. Adhami, ImageSegmentation and Shape Analysis for Road-Sign Detection, IEEETransactions on Intelligent Transportation Systems, vol. 12, pp.83-96, 2011. 7. S. S. Al-Amri and N. V. Kalyankar, "Image segmentation by usingthreshold techniques," arxiv preprint arxiv:1005.4020, 2010. 8. Lim, Y.W., and Lee, S.U.: On the colour image segmentationalgorithm based on thresholding and the fuzzy c-means techniques,pattern Recognit., 23, (3), pp. 935 952, 1990. IJPT June-2016 Vol. 8 Issue No.2 12712-12717 Page 12716
9. Yining Deng, Member, IEEE, and B.S. Manjunath, Member, IEEEUnsupervised Segmentation of Color- TextureRegions in Images and Video, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 23, No. 8, August 2001. 10. Song Wang, Member, IEEE, and Jeffrey Mark Siskind, Member, IEEE, Image Segmentation with Ratio Cut, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 25, No. 6, June 2003. 11. Guo Dong, Member, IEEE, and Ming Xie, Member, IEEE, Color Clustering and Learning for ImageSegmentation Based on Neural Networks, IEEE Transactions on Neural Networks, Vol. 16, No. 4, July 2005. 12. Michael E. Farmer, Member, IEEE, and Anil K. Jain, Fellow, IEEE, A Wrapper-Based Approach to ImageSegmentation and Classification, IEEE Transactions onimage Processing, Vol. 14, No. 12, December 2005. 13. M. A. Gonzalez, G. J. Meschino and V. L. Ballarin, Solving the over segmentation problem in applications of Watershed Transform, Journal of Biomedical Graphics and Computing, vol. 3, no. 3, pp. 29-39, 2013. 14. Lei Zhang, Member, IEEE, and Qiang Ji, Senior Member, IEEE, A Bayesian Network Model for Automatic andinteractive Image Segmentation, IEEE TransactionsonImageProcessing, Vol. 20, NO. 9, September 2011. 15. S. Aksoy, H. G. Akcay, and T. Wassenaar, Automatic mapping of linear woody vegetation features in agricultural landscapes using very high resolution imagery, IEEE Trans. Geosci. Remote Sens., vol. 48, no. 1, pp. 511 522, Jan. 2010. 16. M. Bouziani, G. Kalifa, and D. C. He, Rule-based classification of avery high resolution image in an urban environment using multispectralsegmentation guided by cartographic data, IEEE Trans. Geosci. RemoteSens., vol. 48, no. 8, pp. 3198 3211, Aug. 2010. 17. Q. X. Chen, J. C. Luo, C. Zhou, and T. Pei, A hybrid multi-scale segmentation approach for remotely sensed imagery, in Proc. IEEE Int. Geosci. Remote Sens. Symp., pp. 3416 3419, 2003. 18. Patrick NigriHapp, Raul QueirozFeitosa, Cristiana Bentes, and Ricardo Farias, A Region-Growing SegmentationAlgorithm for GPUs, IEEE Geoscience and Remote Sensing Letters, Vol. 10, No. 6, November 2013. Corresponding Author: Rajesh Kaluri* Email: rajesh.kaluri@vit.ac.in IJPT June-2016 Vol. 8 Issue No.2 12712-12717 Page 12717