# I. Introduction esearch on texture segmentation has been carried out for decades; this is because image analysis, description, illustration, classification, image understanding and restoration are largely dependent on the segmentation results. The texture segmentation plays a vital role in a variety of applications such as medical imaging, textile designs, identification of human faces and expressions, various military applications, remote sensing, robot vision, cartography, identification of vehicles and quality assurance in industries etc. The texture is still a relatively poorly understood phenomenon. It is very easy and natural for human being to understand a texture; it is extremely difficult to define it. That's why many researchers attempted to define texture based on their application and a catalogue of texture definitions is available in literature [1].Texture segmentation [2,3,4,5] break ups an image texture into dissimilar areas depending on a variety of attributes. The attributes can be texture, pixel intensities, color, shape or any other feature of interest according to the particular application. Researchers contributed significantly to the problem of image segmentation in the literature [6,7,8,9,10,11]. Color is one of the important attribute of the texture and there are many segmentation schemes that are based on color [10,12,3,14,15,16,17]. The segmentation methods based on wavelets [18], hidden Markov models [19],multichannel filtering [20], quadtree [21], fractal dimension [22], feature smoo thing [23], split-and-merge methods [24], autoregressive models [25], pyramid node linking [26], local linear transforms [27], Markov random field models [28], and selective feature smoothing with clustering [29] are proposed in the literature. The above methods [18][19][20][21][22][23][24][25][26][27][28][29] attained fine results for texture like mosaics (a tiny set of fine-grained texture); and failed to achieve a precise segmentation for natural texture images. The present paper considered Brodtaz and other natural textures and implemented segmentation on wavelet based images using morphological and thresholding techniques to obtain better results. Edge-based [30,31], region-based [32] and pixel-based segmentation [33] methods are also popular in the literature. Region-based segmentation can identify partitions in a given image. The region based methods [34,[41][42][43] arepopular in literature. The segmentation methods based on normalized cuts are also proposed [35][36][37][38][39][40] and among these, the multi scale normalized cut approach [39] obtained a precise segmentation. The histogram based [44][45][46][47][48] methods, fall in to pixel based segmentation approaches. Thetextonor shape based methods [49][50][51][52] also attained good results. The exactness of segmentation method is highly dependent on i) Type of textures ii) The type of attributes considered iii) The way the attributes are evaluated (global, local or region wise etc.). This indicates that segmentation methods are application dependent. The present paper initially decomposes the texture images using wavelet transforms. The segmentation scheme is applied on the decomposed image and quality assessment parameters are evaluated. The present paper is organized as follows: The section 2 describes the related work. The section 3 and 4 describes the proposed method and results and discussions respectively. The section five describes the conclusions. other applications. Mathematical morphology refers to image components like topology, shape, connectivity etc. The morphological operations are very simple, easy to understand, compute and analyze because they are basically derived from algebraic operators and it is proposed by Matheron and Serra and it is an extension of Minkowski's set theory [53], [54]. Morphology is gained much attention in solving and analyzing image processing problems related to geometrical variations and aspects of the image, whereas most of the nonmorphological image processing methods are mostly unsuccessful in this aspect. These methods have additional advantages in dealing with textures, because the texture is basically nonlinear in nature. The morphology basically deals with shape or topological properties of objects. That is the reason they are most significant in segmentation problems. Using morphological operations one can easily represent or capture the various dissimilarities between geometrical properties such as size, connectivity, shape, which are considered as essential feature parameters that are basically needed to partition or segment an image texture. Many researchers used morphological operations extensively in various computer vision and pattern recognition applications like: preprocessing, boundary detection, removal of noise, image segmentation, image enhancement, image smoothening, image understanding and analysis of images. The main reason for the popular usage of mathematical morphology in image processing is they are based on dilation and erosion operations, which can be implemented in binary and gray leveldomains. # b) Gray Value Morphological Processing The dilation operation, in gray level is given by equation 1. The images grow in size by dilation. The erosion is given by the following equation 2. The image size is reduced in size, based on the specifications of the structure element (SE). The morphological gray level opening and closing are defined in equations 3 and 4. ???????????????? = max [?? ,??]???? {??[?? ? ??, ?? ? ??]} = max ?? ??(1)?????????????? = min [?? ,??]??? {??[?? ? ??, ?? ? ??]} = min ?? (??) (2) ?????????????? ? ?? ?? (??, ??) = max ?? ? min ?? (??) ? (3) ?????????????? ? ?? ?? (??, ??) = min ?? ?max ?? (??)? (4) Where P and Q are the original image and structuring element. The Structuring element Q contains fixed number pixels, which are bounded and convex in nature. The erosion followed by dilation is called morphological opening. In opening the erosion of an image removes all structures that cannot fit inside. Further shrinks all other structures. Then by dilating the result of the erosion with the same structuring element, the structures that are survived by the erosion (were shrunken, not deleted) will be restored. Opening generally smoothes the contour of an image, splits slim isthmuses, and overcomes from thin protrusions affect. The closing smooth the image removes minute holes, and fills gaps in the contour. This retains the uniformity of a local region. # III. Methodology To derive precise segmentation the present paper initially converts the color image in to gray level image using HSV color quantization. The present paper then convert the gray level image into discrete wavelet transform (DWT) using Harr wavelet transform. # a) Wavelet transforms The mathematical function that is used to divide a given function into components of different frequency is called wavelet transform. And each component is studied by wavelets with a resolution that matches its scale. An image signal is passed through an analysis filter bank followed by a decimation operation and analyzed in wavelet transforms. This filter bank consists of a low pass and a high pass filter at each decomposition stage. The low pass filter, corresponds to an averaging operation. The low pass filters extracts the coarse information of a signal. The high pass filter extracts the detail information of the signal and it represents to a differencing operation. The image will be divided i.e., decomposed into four sub-bands i.e. denoted by low-low (LL), high-low (HL), low-high (LH) and high-high (HH). The LH1, HL1 and HH1 sub bands correspond to the detail images i.e., finest scale wavelet coefficients. The LL1 sub-band corresponds to approximation image (coarse level coefficients). By decomposing LL1 sub band alone, the next coarse level of wavelet coefficients will be obtained and they are denoted as LL2, LH2, HL2, and HH2. Similarly, to obtain further decomposition, LL2 will be used. The features obtained from these DWT transformed images are useful for texture analysis, namely segmentation. The # b) Identification of interior regions by Top-hat transform Initially the image is converted in to DWT image. To recognize details of interior regions and to improve the contrast of the image texture, morphological treatment is given by using Top Hat transform (THT). This paper used THT, instead of Histogram Equalization (HE) to deal with the natural textures with high and low mean brightness values; The present paper also overcomes the enhancement disadvantages that arises from the global content of the image. HE may produce over enhancement and saturation artifacts [55,56]. That is why researchers are focusing on upgrading of the conventional histogram equalization. To overcome this present paper derive contrast enhancement based on THT of morphology. The THT is used to extract image features. There two types THTs: white top hat transforms (WTH) and black top hat transformation (BTH). The morphological WTH is applied to take out bright or over intensity features of texture mage. It is derived from equation 5. P WTH = P ? (P ? Q) (5) To extract the darker features of image texture the BTH is used and it is derived by equation 6. P BTH = (P ? Q) -P(6) To increase the contrast between the black and white regions of image the present paper derived a new contrast enhancement using MM operations as given below. P CE = P +P WHT -P (7) where, P is the original image. P CE is the final enhanced image. P WHT represents the extracted white image regions. P b represents the extracted black image regions. To derive the uniform local regions of the texture image, closing operation is applied in this paper. This has connected the objects that are nearer to each other and it has filled small gaps. # c) Thresholding by Otsu method To set upwell defined boundaries in the texture image, Otsu thresholding is applied in this paper. One of the important steps in segmentation is thresholding. Thresholding divides the texture image in to two or more units. Threshold will be chosen based on intensity attribute of the objects, sizes of the objects, number of different types of objects appearing in an image etc. One should have proper knowledge about the images and the application to choose the threshold. The Otsu method [57] is based on discriminate analysis. This method [57] selects the threshold by limiting the withinclass variance of the two groups of pixels separated by the thresholding operator. A measure of region homogeneity is variance. The OTSU thresholddoes not depend on modeling the probability density functions and it is based on a bimodal distribution of gray-level values. The OTSU threshold operation performs the division of image pixels into two classes C0 and C1 (e.g., objects and background) at gray level. # IV. Results and Discussions The performance of the texture segmentation schemes can be assessed by subjective evaluation, supervised and non-supervised evaluation: Comparing the approximate segmentation results with various other segmentation approaches is called subjective evaluation; The segmented output, is compared with the original image in supervised evaluation. The above two assessments are impractical and not popular because they are not automatic and requires human interaction. Year 2017 ( ) # F In "unsupervised approaches" [58,59] comparison with ground truth or original images is not required. And they take less time in evaluating the performance of the segmentation method. The proposed segmentation scheme is assessed by unsupervised parameters like: Discrepancy, Entropy, Standard deviation, internal region contrast as given below. The value of these indicates the following: If the value of discrepancy is high, then it indicates a better segmentation. Using entropy value one can recognize the Over segmentation and under segmentations. Over segmentation will be resulted if entropy value less than 1 and if it is above 1.5 then represents under segmentation. A better segmentation is estimated with lower values of standard deviation. Region uniformity should not be disturbed while segmenting. If the segmented image results a low internal contrast then it indicates a high uniformity. ?????????????????????? = ? ? (??(??, ??) ? ??(??, ??)) ?? ??=1 ?? ??=1 (8) Where A(r,c) and B(r,c) represents the gray level original and segmented image. Entropy of an image is given as ?????????????? = ? ? ? ??(??, ??)?????????(??, c)? ?? ?? (9) Where B is segmented image Standard deviation of a given vector is expressed as Standared deviation S = ? 1 n?1 ? (x i ? x ?) 2 n i=1 ? 1 2 (10) Where x i andx ? are the value of vector and average of all values. Internal region contrast is defined as Internal region contrast I j = 1 S j ? max?cont(??, t), t ? s?R j N???Rj (11) Where N(s) is the neighborhood and contrast (s,t) = |C x (s) -C x (t)| is the contrast of pixel 's' and 't'. The region uniformity is measured by internal contrast, I j . The I j represents "region average Max Contrast". The present method is tested on Brodtaz [60], Oxford flowers [61], Wang [62] and standard images from Google (Lena, Camera man, House, Mandrill, and Ship) [63]. The experiments are carried out by considering 100 images from each data base, thus it results a total of 400images.There are 1,000 natural images in Wang database and these images are selected manually from Corel stock photo database. These images are divided into 10 categories (each category consists 100 images). There are 17 classes (80 images per class) in Oxford flower database. There are 112 texture images in the Brodatz album with different background intensities. The proposed method is compared with three existing methods ISLGHEM [64], automatic thresholding method [65], wavelet based watershed method [66] and MULBP method [67]. To show the performance, the proposed integrated DWT segmentation scheme is applied on input images and results are shown in Fig. 2 The present paper evaluated the above statistical parameter or unsupervised approaches on the proposed and the other methods. The results are di splayed in Table 1, 2, 3 and 4 and also plotted in Fig. 6 to Fig. 9. The proposed method is compared with the existing methods [64,65,66,67]. A high discrepancy rate is noted for all considered textures (14.56 to 15.66) except standard images of Google on the proposed method, which has shown low discrepancy rate of 12.19, and resulted an average value of 14.5.The average value of entropy, standard deviation and internal region contrast for the proposed methods are 1.26, 1.75 and 0.91 respectively, which clearly indicates a good segmentation. The proposed method resulted high discrepancy value and the low standard deviation over the other methods (Fig: 6 and Fig. 9). This reflects a better segmentation of the proposed over the existing methods. The low value of internal contrast and a decent value of entropy also indicate the high performance of the proposed method over the existing methods. # V. Conclusions The morphological operations covered the small holes with intensities and borders are connected for a better segmentation. The Otsu threshold established local boundaries efficiently and provided better contrast and also removed the unwanted local scenes of the image. The whole process of segmentation is automatic and requires no supervision. The present paper improves the contrast of sharp details in light and dark areas. The present method is experimented on the four standard databases namely Brodatz, Wang, Oxford flowers and standard images. The present method attained a good segmentation on the four datasets: however the proposed method have shown significantly high performance on Brodatz database textures when compared with other standard datasets; followed by Wang and oxford datasets. Further no over and under segmentation is reported by the present method on the considered databases. The present method is simple and suitable to real time applications because it achieved good segmentation with three basic steps. 0 0.5 1 1.5 2 2.5 3 3. 5 4 Proposed method MULBP method [67] Automatic threshold method [65] ISLGHEM method [64] Wavelet based watershed method [ 1![Figure 1: Shows the 1-level, 2-level and 3-level wavelet transformed image.](image-2.png "Figure 1 :") 1![Figure 1(a) Original Brodatz Texture (b) Level-1 Wavelet Transformed Brodatz Texture (c) Level-2 Wavelet Transformed Brodatz Texture (d) Level-3 Wavelet Transformed Brodatz Texture.](image-3.png "Figure 1 (") 2345![Figure 2: The segmented output Brodatz texture images by the proposed method.](image-4.png "Figure 2 :Figure 3 :Figure 4 :Figure 5 :") 8![Figure 8: Comparison graph of proposed and existing methods in terms of Average internal region contrast over considered databases](image-5.png "Figure 8 :") 1Year 201712Volume XVII Issue I Version I( ) FGlobal Journal of Computer Science and TechnologyName of the data base Brodatz texture Wang Oxford flower Standard image Average of all databasesProposed method 15.66 15.64 14.56 12.10 14.49ISLGHEM [64] ATM [65] WWM[66] MULBP method [67] 12.56 13.68 11.20 14.3 11.3 13.2 10.3 14.1 11.4 12.3 10.6 13.2 8.3 9.6 7.8 10.3 10.37 11.67 9.65 12.42© 20 7 Global Journa ls Inc. (US) 1 2Name of the data baseProposed methodISLGHEM [64]ATM [65]WWM [66] MULBP method[67]Brodatz texture1.202.352.101.981.4Wang1.213.13.42.61.28Oxford flower1.362.73.12.61.44Standard image1.3022.31.81.35Average of all databases1.262.4752.852.2251.36 3Name of the data baseProposed methodISLGHEM method[64]Automatic threshold method[65]Wavelet watershed based method[66]MULBP method[67]Brodatz texture database1.53.83.33.62.1Wang database1.984.44.13.82.7Oxford flower database1.864.43.54.22.1Standard image database1.694.33.542.6Average of all databases1.754.33.853.92.55 4Year 201714Volume XVII Issue I Version I ( )Average standard deviation value0 1 2 3 4 5Proposed methodMULBP method[67] Proposed and existing methods Automatic threshold method[65] ISLGHEM method[64]Wavelet based watershed method[66]Global Journal of Computer Science and TechnologyName database of Brodatz texture the database Wang database Oxford flower database Standard image databaseProposed method 0.82 0.89 0.86 1.1 Wang database ISLGHEM method[64] 1.20 1.8 1.5 2.8 Broadtaz texture database Automatic threshold method[65] 1.68 1.4 1.8 2.6 Oxford flower database Wavelet watershed method[66] based 2.3 1.9 2.6 3.4 Standard image databaseMULBP method[67] 1.13 0.90 1.10 1.52Average databasesofall0.912.1751.952.6751.28 © 2017 Global Journals Inc. (US) © 20 7 Global Journa ls Inc. (US) Year 2017 ( ) F © 20 7 Global Journa ls Inc. (US) 1 * Texture Based Segmentation using Statistical Properties for Mammographic Images .HKekre Gharge Int. Journal of Advanced Computer Science and Applications(IJACSA) 1 5 Nov 2010 * Towards Fully Automatic , Image Segmentation Evaluation LutzGoldmann TomaszAdamek PeterVajda Advanced Concepts for Intelligent Vision Systems Lecture Notes in Comp. Science 2008 5259 * Texture experiments with Meastex and Vistex benchmarks Singh Sharma Proc. Inter Inter * Conf. on Advances in Pattern recog Lecture Notes in Computer Science 2013 * Supervised texture segmentation using wavelet transform Wang Zhang Proc. of Int. Conf. on Neural Networks and Signal Processing of Int. Conf. on Neural Networks and Signal essing 2003 2 * Segmentation of the Breast Region in Mammograms using a Rule-Based Fuzzy Reasoning Algorithm MWirth DNikitenko JLyon GVIP Journal Special Issue on Mammograms 2007 * A new texture segmentation approach for medical images VijayaKumar ISakakezia Santiprabha International Journal of Scientific & Engineering Research (IJSER) 2229-5518 4 1 2013 * A colortexture based segmentation method to extract object from background ISakakezia VSantiprabha Vijayakumar International Journal Image, And Graphics And Signal Processing 2074-9082 5 2013 IJIGSP) * A new texture based segmentation method to extract object from background M. JosephPrakash VVijayakumar Global Journal of Computer Science And Technology Graphics & Vision (GJCST) 0975 -4350 12 15 2012 * Texture segmentation methods based on combinatorial of morphological and statistical operations VKumar BEswarReddy ARao US NRaju Journal of multimedia JMM * Pipeline implementation of new segmentation based on cognate neighborhood approach VKumar ARao US NRaju BEswarReddy International Journal of Computer Science (IJCS) 1552-6607 35 1 2008 Science publications * Efficient DIS Based Region Growing Segmentation Technique for VHR Satellite Images ICGST-GVIP AnjaliAshwinikunte Bhalchandra Journal 10 3 August 2010 * A Vision Chip for Colour Segmentation and Pattern Matching RECummings PPouliquen MALewis EURASIP Journal on Applied Signal Processing 7 2003 * Fuzzy C-means clustering algorithm with a novel penalty term for image segmentation YYang CZhen PLin Opto-Electronics Rev 13 4 2005 * Volterra Filter for color image segmentation MBMeevathi KRajesh Inter. Journal of Computer Science and Engineering 2 1 2008 * A hierarchical approach to color image segmentation using homogeneity HDCheng YSun IEEE Transaction on Image Processing 9 12 2000 * Color Medical Images Watermarking, Based Neural Network Segmentation HSeddik EBen Braiek Medical Image Processing 2006 * Frame representations for texture segmentation ALaine JFan IEEE Trans. Image Processing 5 1996 * Morphological Multi-scale Stationary Wavelet Transform based Texture Segmentation Jp Mosiganti VVijayKezia Kumar International Journal of Artificial Intelligence & Applications (IJAIA) 2074-9082 8 1 2014 * Unsupervised texture segmentation using multichannel decomposition and hidden Markov models J.-LChen AKundu IEEE Trans. Image Processing 4 1995 * Learning texture discrimination rules in a multiresolution system HGreenspan RGoodman RChellappa CHAnderson IEEE Trans. Pattern Anal. Machine Intelligence 16 1994 * A quad-tree approach to image segmentation which combines statistical and spatial information MSpann RWilson Pattern Recognition 18 1985 * Texture segmentation using fractal dimension BBChaudhuri NSarkar IEEE Trans. Pattern Anal. Machine Intelligence 17 1995 * Unsupervised texture image segmentation using feature smoothing and probabilistic relaxation techniques JYHsiao AASawchuk Computer Vision Graphics Image Processing 48 1989 * Segmentation by texture using a co-occurrence matrix and a split-and-merge algorithm PCChen TPavlidis Comput. Graphics Image Processing 10 1979 * Texture classification and segmentation using multi resolution simultaneous autoregressive models JMao AKJain Pattern Recognition 25 1992 * Image segmentation by texture using pyramid node linking MPietika¬ Inen ARosenfeld IEEE Trans. Systems Man Cybernet 11 1981 * Multiresolution feature extraction and selection for texture segmentation MUnser MEden IEEE Trans. Pattern Anal. Machine Intelligence 11 1989 * Unsupervised texture segmentation using Gabor filters AKJain FFarrokhnia Pattern Recognition 24 1991 * A Markov Random Field model-based approach to unsupervised texture segmentation using local and global statistics CKervrann FHeitz IEEE Trans. Image Processing 4 6 June 1995 * A Survey on Image Segmentation through Clustering Algorithm MLalitha1 MKiruthiga2 CLoganathan3 International Journal of Science and Research 2319-7064 2 2 February 2013 * Discriminative sparse image models for class-specific edge detection and image interpretation JMairal MLeordeanu FBach MHebert JPonce Proc. 10th European Conf. Computer Vision 10th European Conf. Computer Vision 2008 * Automatic image segmentation by dynamic region merging BPeng LZhang DZhang IEEE Trans. Image Process 12 2011, 20 * Contour detection and hierarchical image segmentation PArbelaez MMaire CFowlkes JMalik IEEE Trans. Patt. Anal. Mach. Intell 5 2010 * Practical handbook on image processing for scientific and technical applications BJähne 2004 CRC Press 15 2nd Ed * Normalized cuts and image segmentation JShi JMalik IEEE Trans. Patt. Anal. Mach. Intell 22 8 2000 * Contour and texture analysis for image segmentation JMalik SBelongie TLeung JShi Int. J. Comput. Vis 1 2001 * Multiclass spectral clustering SXYu JShi Proc. Int. Conf. Computer Vision Int. Conf. Computer Vision October 2003 * I-247-I-254. normalized-cut segmentation for spinal MRI SXYu Proc. IEEE Conf. on Comput. Vision Pattern Recognition IEEE Conf. on Comput. Vision Pattern Recognition June 2004. 2009 1 Segmentation using multiscale cues * Segmentation induced by scale invariance SXYu Proc. IEEE Conf. on Comput. Vision Pattern Recognition IEEE Conf. on Comput. Vision Pattern Recognition June 2005 1 * Spectral segmentation with Multiscale graph decomposition TCour FBenezit JShi Proc. IEEE Conf. on Computer Vision Pattern Recognition IEEE Conf. on Computer Vision Pattern Recognition June 2005 2 * Acceleration strategies for Gaussian mean-shift image segmentation MACarreira-Perpinan Proc. IEEE Conf. on Computer Vision Pattern Recognition IEEE Conf. on Computer Vision Pattern Recognition June 2006 1 * Unified mean shift segmentation and graph region merging algorithm for infrared ship target segmentation W3tao HJin JLiu Opt. Eng 46 12 2007 * An adaptive mean-shift framework for MRI brain segmentation AMayer HGreenspan IEEE Trans. Med. Imag 28 8 2009 * Fast image segmentation based on K-means clustering with histograms in HSV color space TWChen YLChen SYChien Proc. IEEE Int. Workshop on Multimedia Signal Processing IEEE Int. Workshop on Multimedia Signal essing October 2008 * Color image segmentation using histogram Thresholding -Fuzzy C-means hybrid approach KSTan NA MIsa Patt. Recogn 1 2011 * Color image segmentation using adaptive unsupervised clustering approach KSTan NA MIsa WHLim Appl. Soft Comput 13 2012 * Novel initialization scheme for Fuzzy C-Means algorithm on color image segmentation KSTan WHLim NA MIsa Appl. Soft Comput 13 2013 * An efficient method for segmentation of images based on fractional calculus and natural selection PGhamisi MSCouceiro JABenediktsson NMFerreira Expert Syst. Appl 39 16 2012 * Texton-based segmentation of retinal vessels ADonald UmasankarkandaswamyAdjeroh Journal of Optical Society of America 24 5 May 2007 * Texture image segmentation using a new descriptor and mathematical morphology SamirIdrissisidiyassine Belfkih Int. Arab Journal of Information Technology 10 2 March 2013 * Innovative segmentation approach based on LRTM SKezia IShantiprabha Vijayakumar Int. Journal of Soft Computing and Engg 2 5 Nov. 2012 * Image Analysis and Mathematical Morphology JSerra 1982 Academic London, U.K * Introduction to mathematical morphology JSerra Comput. Vis., Graph., Image Process 35 1986 * Image coding using wavelet transform MAntonini MBarlaud PMathieu IDaubechies IEEE Trans. Image Process 1 2 1992 * Fast Computation of Multiscale Morphological Operations for Local Contrast Enhancement YDerong ZYuanyuan LDongguo Proceedings of the 2005 IEEE the 2005 IEEE * A threshold selection method from graylevel histograms Otsu IEEE Trans. Sys., Man., Cyber 9 1979 * An Early Fire Detection Method Based on Smoke Texture Analysis and Discrimination DongY HCui .EZhou Journal Congress on Image and Sig. Proc 2008 * A new texture segmentation method based on the fuzzy Cmean algorithm and statistical features MounirSayadi Lotfitlig FarhatFnaiech Applied Mathematical Sciences 1 60 2007 * A Photographic Album for Artists and Designers PBrodatz Textures 1999 Dover New York, NY, USA * A New Texture Segmentation Approach for Medical Images Saka Kezia .IDr Santiprabha .VDr Vijayakumar International Journal of Scientific & Engineering Research 4 1 January-2013 * Image Segmentation Method Using Thresholds Automatically Determined from Picture Contents OscalT CYuan Been Chen Chen Eurasip journal on image and video processing 2009. 2009 * Wavelet-based Watershed for Image Segmentation Yu-HuaChai Li- ShunLu LeiTian Proc. of the 6th World Congress on Intelligent Control and Automation of the 6th World Congress on Intelligent Control and AutomationDalian, China June 21 -23, 2006 * A Segmentation scheme based on uniform LBP and morphological approach, Graphics SGAnuradha KKaribasappa BEswarReddy Vision and Image Processing (ICGST-GVIP) 2016 16