International Journal of Engineering Trends and Technology (IJETT) – Volume 12 Number 8 - Jun 2014 Image Segmentation Based on the Edge Detection with Snakes Contouring and Ant Colony Optimization Kirandeep kaur-MTech(IT) Department of Computer Science & Technology, Lovely professional University, Phagwara, Punjab, India Abstract--Image segmentation means dividing an image into two parts which are loosely connected. Segmentation is on the basis of Edge-Based and Region-Based Techniques. Edge detection filters the important features in an image from the edges and detects the edges where there are some changes of intensity. It has many applications and based on lots of image analysis. Medical images in image processing are one of the most interesting and challenging topics in the systematic investigation field. Image segmentation is to filter out many features of an image i.e. used for image analysis. MRI (Medical Resonance Image) has an important role in medical diagnostics. The proposed technique is combination of “snakes” contouring and Ant colony optimization techniques. Snake is a multistage process and ACO is probabilistic process which takes the decision through the movement of number of ants. In the experimental results, it’s provided to demonstrate the quality performance of the proposed technique. images inside the body. This technique is used in every hospital for medical diagnosis. Snakes find the contour that best approximates the perimeter of an object. Snakes are active contour models. It represents some salient features of MRI as a parametric curve. It is like a rubber band. User makes any curve close to the object boundary and then starts moving towards the desired object boundary. In ACO, Pheromone is a parameter which determines the intensity of the trail or probability. II. SEGMENTATION The segmentation might be gray-level, texture, colour, depth or motion etc.In segmentation, gray level images can provide useful information of the surfaces where some action occurs. Example of Segmentation as: Keyword-- Segmentation, Edge detection, Ant colony optimization, Snakes contouring. I.INTRODUCTION Segmentation has various techniques such as Edge-Based, Region–Based and Adaptive thresholding based segmentation. Image segmentation divides an image into set of area [3]. It includes various applications in the field of medical imaging, satellite imaging, movement detection; surveillance etc.The main objectives of segmentation are (i) Break down the image into regions. (iii)Change of representation of an image. Edges occur when we have a boundary between two regions in an image. Edge detection is also known as “Extraction of edges”. This detects boundary between background and objects in the image where the brightness of image changes abruptly. This includes various methods such as Roberts, Prewitt, Sobel, Canny, Laplacian of Gaussian (LOG) etc [4]. MRI investigates anatomy and function of the body in health and disease. And its scanners use strong magnetic fields and radio waves to make ISSN: 2231-5381 Fig.1Example of segmentation Complexity of the images is a challenging problem in medical imagery in segmentation i.e. Brain tissue [7].For Brain’s segmentation, we can use the two techniques such as image processing and model-based techniques [7]. Our hybrid method follows the first image processing steps via a threshold step which removes small connections between the brain and surrounding tissue. Then model-based used to take out the eyes and other features of other nonbrain and recover some of the http://www.ijettjournal.org Page 378 International Journal of Engineering Trends and Technology (IJETT) – Volume 12 Number 8 - Jun 2014 terminated tissue by morphological dilation, a last refinement of the brain contour (i.e.”Snakes” active contour algorithm). -1 -1 -1 III.EDGE DETECTION Edge detection is used to reduce data of an image and produce image outlines. Edges are those pixels where intensity of image abruptly or suddenly changes. It simply characterizes boundaries of objects in image. Geometric events (surface discontinuities, depth discontinuities, color discontinuities, texture discontinuities) and non-geometric events (illumination changes, secularities, shadows, inter reflections) are the main reasons for the intensity change. To determine the edge strength and direction of maximum intensity change at one point and this is perpendicular to the direction of the gradient vector at that point. In the following figure, we are describing edge strength and edge normal. Edge normal is unit vector in which intensity change is maximum and the vector which is perpendicular to this edge normal vector is known as edge direction [9]. 1 0 0 1 1 1 1 0 -1 1 0 -1 1 0 -1 Robert edge detector: Robert method detects the edges using derivative. When the gradient of the image is maximum then it returns edges at those pixels. 1 0 0 -1 0 -1 1 0 Prewitt edge detector: Prewitt depends upon the idea of central difference only in one direction and less susceptible to noise [4]. -1 -1 -1 1 0 0 1 1 1 1 0 -1 1 0 -1 1 0 -1 IV.PROPOSED METHOD (A)Snakes contour: In this algorithm, we use three-stage method. Firstly, we eliminate the background noise (f)= Fig.2Edge normal and Edge direction The following methods are using IST derivative and 2nd derivative which is defined as gradient operator and gradient direction [10] as: -1 -1 -1 1 0 0 1 1 1 1 0 -1 1 0 -1 1 0 -1 Sobel operator: It detects the edges along horizontal-axis (180o) and Vertical-axis (90o). Sobel depends upon convolving with an integer valued filter. ISSN: 2231-5381 , here f is the noise intensity and is the standard deviation of white noise then generate initial brain mask which contains the three-steps (i)Nonlinear anisotropic diffusion (ii)automatic threshold (iii) Mask refinement. In third step, we refine the mask which is using the “snakes” algorithm and it is defined as determining the boundary between the brain and the intracranial cavity. It makes formless contours through the initial brain mask’s perimeter to lock onto edge of the brain [7].A particular active contour is a collection of n points as: = { 1, … … … … . , =( , Canny Edge detector: Canny edge detector is a multistage process. It detects a wide range of edges in images. In this, we use Gaussian convolution to smoothen the image and then highlight the regions of image using first derivative. − } (1) ) = {1, … … … . . , } (2) Here each point vi, an energy matrix E (vi) is expressed as: E(vi)= (3) ( )+ ( )+ ( )+ ( ) ( ) is continuity, ( ) is balloon Here, ( ) is an intensity energy force, ( ) is a force, gradient energy function. Each vi moved to the point where minimum energy in its neighborhood which is the smallest element in E (vi). http://www.ijettjournal.org Page 379 International Journal of Engineering Trends and Technology (IJETT) – Volume 12 Number 8 - Jun 2014 Fig.4(c) Fig.4 (d) Fig.3.Refinement of intracranial contour (a) The contour defined by the perimeter of initial brain mask. (b) The intracranial contour detected using the active contour model algorithm. (B)Ant colony optimization: ACO-based image edge detection defined as a number of ants to move for making a pheromone matrix. The movement of ants goes in the direction that you wanted by the local variation of intensity values in the image. In the proposed algorithm, we start from the initialization process then N iterations to construct the pheromone matrix through construction process and the last one is the decision process which determines the edge. We are using the following four functions and are expressed as [8]: F(x) = if x>=0 (4) F(x) = x 2 if x>=0 (5) F(x) = sin 0 F(x) = sin 0 else (6) 0=<x=< else (7) Here, is a parameter, figure (a), (b), (c), (d), (e) present a different results of image cameraman and it show as: Fig.4 (a) Fig.4.(a)Original Image function’s equation (4) function’s equation (5) function’s equation (6) function’s equation (7) (b)Proposed ACO (c) Proposed ACO (d) Proposed ACO (e) Proposed ACO An ACO and Snakes contour based edge detection has been successfully developed. Segmentation of MRI brain images help in the accurate detection of the region of interest. In the proposed approach, both provide a superior performance [5]. Snakes contour detects the intracranial boundary which has proven on research MRI data sets provided from various scanners. In ACO is performed as to refine the edges information( i.e. extracted by ACO) [5].Furthermore for future research work, the parallel Ant colony optimization algorithm can be reduce the computational load of this experiment[2].Automatic brain segmentation to the problem of lung segmentation for lung diseases(cystic fibrosis and emphysema where we needed the volume of the lungs. Early results with MR images are promising and can be continued. Fig.4 (b) ISSN: 2231-5381 with with with with V. CONCLUSION AND FUTURE WORK 0=<x=< Fig.4 (e) http://www.ijettjournal.org Page 380 International Journal of Engineering Trends and Technology (IJETT) – Volume 12 Number 8 - Jun 2014 REFERENCES [1]Rafael C. Gonzalez and Richard E. Woods (2008) Digital Image Processing, Pearson Education p. 954. [2]M.Randall and A.Lewis, “A parallel implementation of ant colony optimization “, Journal of parallel and Distributed Computing, vol.62, pp.14211432, Sep.2002. [3]G.Evelin Sujji, Y.V.S.Lakshmi, G.Wiselin Jijji “MRI brain image segmentation based on thresholding”, International Journal of Advanced Computer Research, pp.97-101, March.2013. [4]Prof.Bhombe D.L, Ugale M.B, “A hybrid approach to edge detection”, International Journal of Engineering And Computer Science, pp.1395-1399, April.2013. [5]H.Nezamabadi-pour, S.Saryazdiand E.Rashedi, “Edge detection using an ant algorithm”, soft computing, vol.10, pp.623-628, May.2006. [6]M.Stella Atkins and Blair T.Mackiewich, “Fully automatic segmentation of the brain in MRI, IEEE Transaction on Medical Imaging, vol.17, no.1, February.1998. [7]T.Kapur, W.e.LGrimson, W.M.Wells III and R.Kikinis, “Segmentation of brain tissue from magnetic resonance images”, Med.Imag.Anal. vol.1.no.2,1996. [8]Jing Tian, Weiyu Yu and Shengli Xie, “An Ant Colony Optimization algorithm for image edge detection”, IEEE Congress on Evolutionary Computation, 2008, pp.751-756. [9]Trucco, Chapter 4 AND et al... Jain, Chapter5, pp1-29 [10]Djemel Zoiu and Salvatore Tabbone, “Edge detection techniques-An Overview”, Department de math et informatique university de SherbrookQuebec, Canada, pp.141. ISSN: 2231-5381 http://www.ijettjournal.org Page 381