# Introduction olor segmentation is an essential, critical, and preliminary process in a lot of vision-based tasks such as object recognition, visual tracking, human-computer interaction (HCI), human-robot interaction (HRI), vision-based robotics, visual surveillance, and so forth, because color is an effective and robust visual cue for characterizing an object from the others. Recently, there has been growing interest in the insect segmentation, which aims at detecting insects in an image. However, color segmentation is not robust enough to deal with complex environments. Especially, changing illumination condition and complex background containing surfaces or objects with similar colors to a target are the major problems that limit its applications in practical real world. The former changes the characteristics of a color and the latter results in increasing false positive pixels. Generally, it is known that robustness in color segmentation is achieved if a Author ? : M.Joseph Prakash, Associate Professor, IT Dept., GIET, Rajahmundry, Having 16 years of experience. He is pursuing his Ph.D from JNT University, Kakinada in Computer Science under the guidance of Dr. V. Vijaya Kumar. E-mail : mjosephp7@gmail.com Author ? : Dr.V.Vijayakumar, Dean Computer Sciences, Head-Srinivasa Ramanujan Research Forum, GIET Rajahmundry, India. E-mail : vakulabaranam@hotmail.com color space efficiently separates chrominance and luminance in a color image and a plausible model of chrominance distribution is used [1]. Such chrominance information may be utilized to locate possible regions of a color in a color space and additional features may be adopted in order to validate the hypothesis. Ruizdel-Solar and Verschae [2] have compensated for their color segmentation methods with additional features to obtain more valuable results robust to brightness variations. However, this approach is not included in a pure research area of color segmentation. Dai and Nakano [3] have enhanced skin regions by converting red-green-blue (RGB) color signals to YIQ representation and using the I component, which includes color components from orange to cyan. Fieguth and Terzopoulos [4] have developed a tracking algorithm by heavily relying on a color cue composed of red (R), green (G), and blue (B) color components. Evaluating each component of several color spaces, Gomez [5] has listed top components and made a hybrid color space from those. In above researches, a color is dependent on the intensity of a pixel [6]. In real world cases, however, it is not always possible to control illumination condition. Therefore, many researches have been carried out for invariant detection of a color under illumination variations. Yang and Waibel [7] proposed color histograms in normalized red-green (RG) chromatic color space [6]. McKenna et al. [8] have proposed, however, Gaussian mixture models for the task, which outperform single Gaussian model. Also, color segmentation algorithms have been proposed in order to obtain the robustness toward changes in illumination and shadows by dropping the intensity. Sobottka and Pitas [10] have used chromatic information in hue-saturation-intensity (HSI) color space together with a best-fit ellipse technique to improve robustness of color segmentation. Tomaz et al. [11] have proposed an algorithm for color segmentation in TSL color space. Moreover, various color models have been proposed by using several color spaces such as YCbCr [12], YUV [13], and CIE Lab [14]. Actually, Phung et al. [15] have ascertained that most color space transformations do not bring the assumed benefits. Especially, Jayaram et al. [28] have verified that the best performance of skin color segmentation was obtained in HSI color space, keeping an intensity component. That is, color segmentation is largely unaffected by the choice of a color space. However, segmentation performance degrades when only chrominance channels are used in classification. [28] Cho et al. [29] have proposed a method called an adaptive skin color filter that detects skin color regions in a color image by adaptively adjusting its threshold values. Soriano et al. [30] have demonstrated a chromaticity-based constraint to select training pixels in a scene. In [31], histograms have been dynamically updated, based on feedback from the current color segmentation and prediction of a Markov model to get changing geometric parameters of color distribution and to track those. Adaptive or learning methods for color segmentation dynamically allocated a color through various illumination conditions while those models were adjusted for every image sequences and the adjustment needs heavier computational load to track targets than a static color model. Since there are a number of models for color segmentation as above, Martinkauppi et al. [32] have selected four types of color segmentation and compared their performance each other under significantly changing illumination conditions. Phung et al. [15] have investigated nine different color pixel classification algorithms. It was found for the Bayesian classifier to have higher classification rates than the other tested classifiers. Generally, color image segmentation approaches can be divided into the following categories: statistical approaches, edge detection approaches, region splitting and merging approaches, methods based on physical reflectance models, methods based on human color perception, and the approaches using fuzzy set theory [33], [34].Histogram thresholding is one of the widely used techniques for monochrome image segmentation [35]. As for color images, the situation is different due to the multifeatures [36]. Since the color information is represented by tristimulus R, G and B or some linear/nonlinear transformation of RGB, representing the histogram of a color image in a three-dimensional (3-D) array and selecting threshold in the histogram is not a trivial job [37]. One way to solve this problem is to develop efficient methods for storing and processing the information of the image in the 3-D color space. [38] used a binary tree to store the 3-D histogram of a color image, where each node of the tree includes RGB values as the key and the number of points whose RGB values are within a range centered by the key value. [39] also utilized the same data structure and similar method to detect clusters in the 3-D normalized color space (X, Y, I). Another way is to project the 3-D space onto a lower dimensional space, such as two-dimensional (2-D) or even one-dimensional (1-D). [40] used projections of 3-D normalized color space (X, Y, I) space onto the 2-D planes (X-Y, X-I, and Y-I) to interactively detect insect infestations in citrus orchards from aerial color infrared photographs. [41] Provided segmentation approaches using 2-D projection of color space. [42] Suggested a multidimensional histogram thresholding scheme using threshold values obtained from three different color spaces (RGB, YIQ, and HSI). This method used a mask for region splitting and the initial mask included all pixels in the image. For any mask, histograms of the nine redundant features (R, G, B, Y, I, Q,H,S, and I) of the masked image are computed, all peaks in these histograms are located, the histogram with the best peak is selected and a threshold is determined to split the masked image into two sub regions for which two new masks are generated for further splitting. This operation is repeated until no mask left unprocessed, which means none of the nine histograms of existing regions can be further thresholded and each region is homogeneous. This paper proposed a new color texture based image segmentation approach for insect extraction from complex background. In section II, GLCM was discussed. Automatic OTSU threshold was discussed was given in section III. Methodology of the proposed algorithm was presented in section IV. Results & discussions were given in section V. Finally, conclusions were given in section VI. # II. # GLCM Your Gray level co-occurrence matrix (GLCM) has been proven to be a very powerful tool for texture image segmentation [16,17]. The only shortcoming of the GLCM is its computational cost. Such restriction causes impractical implementation for pixel-by-pixel image processing. In the previous works, GLCM computational burden was reduced by two methods, at the computation architecture level and hardware level. D. A. Clausi et. al. restructures the GLCM by introducing a GLCLL (gray level co-occurrence linked list), which discard the zero value in the GLCM [18].This technique gives a good improvement because mostly GLCM is a sparse matrix where most of its values are equal to zeroes. Thus the size of GLCLL is significantly smaller than GLCM. Then the structure of the GLCLL was improved in [19,20]. Another work is presented in [21] where fast calculation of GLCM texture features relative to a window spanning an image in a raster manner was introduced. This technique was based on the fact that windows relative to adjacent pixels are mostly overlapping, thus the features related to the pixels inside the overlapping windows can be obtained by updating the early calculated values. In January 2007, S. Kiranyaz and M. Gabbouj proposed a novel indexing technique called Hierarchical Cellular Tree (HCT) to handle large data [22]. In his work, it was proved that the proposed technique is able to reduce the GLCM texture features computation burden. GLCM is a matrix that describes the frequency of one gray level appearing in a specified spatial linear relationship with another gray level within the area of ( D D D D ) F 2012 Year investigation [23]. Here, the co-occurrence matrix is computed based on two parameters, which are the relative distance between the pixel pair d measured in pixel number and their relative orientation ?. Normally, ? is quantized in four directions (00, 450,900 and 1350) [23]. In practice, for each d, the resulting values for the four directions are averaged out. To show how the computation is done, for image I , let m represent the gray level of pixels (x, y) and n represent the gray level of pixels ( x±d?0, y±d?1 ) with L level of gray tones where 0 ? x ? M ?1, 0 ? y ? N ?1 and 0 ? m, n ? L ?1. From these representations, the gray level cooccurrence matrix Cm, n for distance d and direction ? can be defined as. ?? ?? ,??,? = ? ? ??{??(??, ??) = ??& ??(?? ± ??? 0 , ?? â??" ??? 1 ) = ??} ?? ?? (1) Where P{.} =1 if the argument is true and otherwise, P {.} =0. For each ? value, its ?0and ?1 values are referred as in the Table 1. In the classical paper [24], Haralick et. al introduced fourteen textural features from the GLCM and then in [25] stated that only six of the textural features are considered to be the most relevant. Those textural features are Energy, Entropy, Contrast, Variance, Correlation and Inverse Difference Moment. Energy is also called Angular Second Moment (ASM) where it measures textural uniformity [23]. If an image is completely homogeneous, its energy will be maximum. Entropy is a measure, which is inversely correlated to energy. It measures the disorder or randomness of an image [23]. Next, contrast is a measure of local gray level variation of an image. This parameter takes low value for a smooth image and high value for a coarse image. On the other hand, inverse difference moment is a measure that takes a high value for a low contrast image. Thus, the parameter is more sensitive to the presence of the GLCM elements, which are nearer to the symmetry line C (m, m) [23]. Variance as the fifth parameter is a measure that is similar to the first order statistical variables called standard deviation [26]. The last parameter, correlation, measures the linear dependency among neighboring pixels. It gives a measure of abrupt pixel transitions in the image [27]. # III. OTSU THRESHOLDING This method, as proposed by [25] is based on discriminate analysis. The threshold operation is regarded as the partitioning of the pixels of an image into two classes C0 and C1 (e.g., objects and background) at grey-level t, i.e., C0 = {0, 1, 2, t} and C1 = {t + 1, t +2,?.L-1}. Let ?2w, ?2B and ?2T be the within-class variance, between-class variance, and the Where ni is the number of pixels with grey-level and n is the total number of pixels in a given image defined as total variance, respectively. An optimal threshold can be determined by minimizing one of the following (equivalent) criterion functions with respect to: Moreover, Pi is the probability of occurrence of grey-level i. For a selected threshold 't' of a given image, the class probabilities w0 and w1 indicate the portions of the areas occupied by the classes C0 and C1. The class means µ0 and µ1 serve as estimates of the mean levels of the classes in the original grey-level image. Moreover, the maximum value of ? , denoted by ?*, can be used as a measure to evaluate the separability of classes C0 and C1 in the original image or the bimodality of the histogram. This is a very significant measure because it is invariant under affine transformations of the grey-level scale. It is uniquely determined within the range 0 ? ? ?1.The low er bound (zero) is obtained when and only when a given image has a single constant grey level, and the upper bound IV. # METHODOLOGY In the first step RGB image is converted to 1976 CIE LUV color space. This color space is approximately perceptually uniform. The perceptual non-uniformity of this color space greatly improves over the CIE XYZ color space. In CIE LUV color space, L* specifies brightness of colors on a scale from 0 to 100, u* specifies color location approximately along the red-green axis with grey located at 0 and v* specifies color location approximately along the yellow-blue axis with grey located at 0.The color corresponding to u*=v*=L*=0 is black and u*=v*=0, L*=100 is white. Each color space has its own appear background and application region. When segmenting a color image, the selection of color space plays a decisive role on the segmentation results. The common color spaces used in color image processing include RGB color space, HSI color space, CIE color space, and so on. At present, the general color digital images are RGB format. RGB color space is based on the theory of three-basic color to build. RGB format is the most basic color space. Other color space models can be obtained through the RGB format conversion. But the RGB color space is not a homogeneous visual perception space, it is not conducive to image segmentation based on color feature. HSI color space uses color characteristics of a direct sense of the three quantities: the brightness or lightness (I), hue (H), saturation (S) to describe the color. This method is more in line with the human eye habits to the description of the color, but the expressed colors are incomplete visual perceived color. One problem with the CIEXYZ colour model is its lack of perceptual balance. Colours which are the same distance from one another are not necessarily perceptually equidistant. In 1976, the CIE proposed the CIE LUV colour model to address this problem. CIELUV is a perceptually uniform colour space. This means that distance and difference can be interchanged as required. If colours A and B are twice as far apart as colours C and D, then the perceived difference between A and B is roughly twice the perceived difference between C and D. The equations for computing CIE LUV assume you have (X, Y,Z) of the colour to convert, and (Xw,Yw, Zw) of a standard white. Given these values, the corresponding LUV colour is: L*=116(Y/Yw)1/3-16,(Y/Yw)>0.01 (10) u*=13L*(u'-u'w) (11) v*=13L*(v'-v'w) (12) u'=(4X)/(X+15Y+3Z) (13) v'=(9Y)/(X+15Y+3Z) (14) u'w =0.2009, v'w=0.4610 (15) L* encodes the luminance or intensity of a given colour, while u' and v' control its chromaticity. The 'v' component is extracted from the transformed LUV color space in the second step. In the third step the extracted v-component is subjected to OTSU thresholding. The two-dimensional OTSU algorithm automatically selects the optimal threshold for segmentation. Because two-dimensional OTSU algorithm not only takes into account the grayscale information of pixels, but also considers the spacerelated information of pixels and their neighborhoods. GLCM is obtained for the otsu thresholded image. For each image and with distance set to one, four GLCMs having directions 0°, 45°, 90° and 135° are generated. The co-occurrence matrix is often correlated with the directions. It is necessary to select more than one direction of gray level co-occurrence matrix for a comprehensive statistical processing. The synthetic gray level co-occurrence matrix of an image can be got by averaging the values of energy matrices in 0 degree, 45 degree, 90 degree and 135 degree. In the fifth step, image negative is applied for better enhancement of insect region. In the last step, morphological closing operation is performed in order to fill small holes. A hole is defined as an area of dark pixels surrounded by lighter pixels or may be defined as a background region surrounded by a connected border of foreground pixels. This process can be used to make objects in an image seem disappear as they are replaced with values that blend in with the background area. This function is useful for image editing, including removal of extraneous details or artifacts. # RESULTS The proposed algorithm is tested on dreamstime image database. In this paper step wise results of insect images are shown in Figs. 2-4 # CONCLUSION The information which the commonly used grayscale images contain is not enough for insect segmentation. The images consisting of insects when collected in nature, the is generally more complex and more close to the target color, so there are some limitations in these conditions if we only use gray level information. However, the color images are able to provide more information. There are colors and color depth information, in addition to its provision of brightness and color images can be expressed by a variety of color space. Therefore, segmentation based on color image can overcome some shortcomings of gray-scale image. In this algorithm both color and texture features are considered. In this method, segmentation speed is faster and without human participation, the segmentation result is also deal. The results show the efficiency of the above algorithm. ![Based Segmentation Method to extract Object from Background](image-2.png "") 1![Fig. 1 : Flowchart of proposed algorithm](image-3.png "Fig. 1 :") 234![Fig. 2 : Step wise results of the proposed algorithm for insect extraction (a) Original image (b) 'v' component (c) OTSU thresholded (d) GLCM output (e) negative (f) final output](image-4.png "Fig. 2 :Fig. 3 :Fig. 4 :") 1 © 2012 Global Journals Inc. (US) * Detection of human faces in complex scene images by use of a skin color model and of invariant Fourier-Mellin moments JTerrillon MDavid SAkamatsu International Conference on Pattern Recognition August 16-20, 1998 2 1350 * Skin detection using neighborhood information JRuiz-Del-Solar RVerschae IEEE International Conference on Automatic Face and Gesture Recognition 2004 * Face-texture model based on SGLD and its applications in face detection in a color scene YDai YNakano Pattern Recognition 29 6 1996 * Color-based tracking of heads and other mobile objects at video frame rates PFieguth DTerzopoulos IEEE International Conference on Computer Vision and Pattern Recognition 1997 * On selecting colour components for skin detection GGomez International Conference on Pattern Recognition .MYang DKriegman NAhuja August 11-15, 2002. 2002 2 Detecting faces in images: a survey * Skin-color modeling and adaptation JYang WLu AWaibel Asian Conference on Computer Vision 1998 2 * Modelling facial colour and identity with Gaussian mixtures SMckenna SGong YRaja Pattern Recognition 31 1998 * Do mixture models in chromaticity space improve skin detection? TSCaetano SDOlabarriaga DA CBarone Pattern Recognition 36 2003 * Extraction of facial regions and features using color and shape information KSobottka IPitas International Conference on Pattern Recognition August 25-29, 1996 3 * Fast and accurate skin segmentation in color images FTomaz TCandeias HShahbaskia Canadian Conference on Computer 2004 * Face detection in color images RLHsu MAbdel-Mottaleb AKJain IEEE Trans. Pattern Anal. Mach. Intell 24 5 2002 * Face detection in complex environments from color images MAbdel-Mottaleb A IEEE International Conference on Image Processing Kobe, Japan 1999 3 * Real-time detection of nodding and head-shaking by directly detecting and tracking the between eyes SKawato JOhya IEEE International Conference on Automatic Face and Gesture Recognition Grenoble 2000 * Skin segmentation using color pixel classification: analysis and comparison SLPhung ABouzerdoum DChai IEEE Trans. Pattern Anal. Mach. Intell 27 1 2005 * A Comparative Study of Texture Measures For Terrain Classification JWeszka CDyer ARosenfeld IEEE Trans. SMC 6 4 April 1976 * A Theoretical Comparison of Texture Algorithms RWConners CAHarlow IEEE 1980 * A fast method to determine co-occurrence texture features DAClausi MEJernigan IEEE Trans on Geoscience & Rem. Sens 36 1 1998 * An advanced computational method to determine co-occurrence probability texture features DAClausi YongpingZhao IEEE Int. Geoscience and Rem. Sens. Sym 4 2002 * Time and space results of dynamic texture feature extraction in MR and CT image analysis AESvolos ATodd-Pokropek IEEE Trans. on Information Tech. in Biomedicine 2 2 1998 * Fast algorithms for texture analysis using co-occurrence matrices FArgenti LAlparone GBenelli IEE Proc on Radar and Signal Processing 137 6 1990 * Hierarchical Cellular Tree: An Efficient Indexing Scheme for Content-Based Retrieval on Multimedia Databases SKiranyaz MGabbouj IEEE Transactions on Multimedia 9 1 2007 * An Investigation of the Textural Characteristics Associated With GLCM Matrix Statistical Parameters ABaraldi FParmiggiani IEEE Trans. on Geos. and Rem.Sens 33 2 1995 * Texture Features For Image Classification RHaralick KShanmugam IDinstein IEEE Transaction 6 1973 * A Threshold Selection Method from Gray-Level Histogram NOtsu IEEE Trans. on System Man Cybernetics 9 1979 * A comparison of spatial feature extraction algorithms for land-use classification with SPOT HRV data PGong JDMarceau PJHowarth Proc.Remote Sensing Environ .Remote Sensing Environ 1992 40 * A tool based on the GLCM to measure the performance of dynamic range reduction algorithms AUkovich GImpoco GRamponi IEEE Int. Workshop on Imaging Sys. & Techniques 2005 * Effect of colorspace transformation the illuminance component, and color modeling on skin detection SJayaram SSchmugge MCShin LVTsap IEEE Computer Society Conference on Computer Vision and Pattern Recognition 2004 * Adaptive skin-color filter KCho JJang KHong Pattern Recognition 34 2001 * Adaptive skin color modeling using the skin locus for selecting training pixels MSoriano BMartinkauppi SHuovinen MLaaksonen Pattern Recognition 36 2003 * Skin color-based video segmentation under time-varying illumination LSigal SSclaroff VAthitsos IEEE Trans. Pattern Anal. Mach. Intell 26 7 2004 * Detection of skin color under changing illumination: a comparative study BMartinkauppi MSoriano MPietikainen International Conference on Image Analysis and Processing September 17-19, 2003 * Color image segmentation: Advances and prospects HDCheng XHJiang YSun JLWang Pattern Recognit to be published * A reviewon image segmentation techniques NPal S Pattern Recognit 26 9 1993 * Adaptive color segmentation-A comparison of neural and statistical methods ELittmann HARitter ; M Arbib IEEE Trans. Pattern Anal. Machine Intell 8 Jan. 1997. Dec.94 IEEE Trans. Neural Networks * Image segmentation techniques RMHaralick LGShapiro Tech. Rep. CVGIP 85 * Scene segmentation by cluster detection in color space BSchacter LDavis ARosenfeld Dept. Comput. Sci Nov. 1975 * Segmentation of chromatic images ASarabi JKAggarwal Pattern Recognit 13 6 1981 * Interactive computer analysis of aerial color infrared photographs SAUnderwood JKAggarwal Comput. Graph. Image Process 6 1977 * An interactive facility for scene analysis research JMTenenbaum TDGarvey SWeyl HCWolf Artif. Intell. Center, Stanford Res. Institute 87 1974 Tech. Rep. * Picture segmentation using a recursive region splitting method ROhlander KPrice DRReddy Comput. Graph. Image Process 8 1978