Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.

com Volume 2, Issue 3, May June 2013 ISSN 2278-6856

Preferential Image Segmentation Using Level Set Methods Based On Color, Boundary and Texture
Sasi Kumar .S
Verizon Data Services India pvt.ltd, Chennai

ABSTRACT: The major key point of preferential image

segmentation, is to preferentially segment objects of interests from an image and ignore the remaining portions of the image for the real world applications. These methods, however, tend to classify or parse a whole image instead of meaningful objects. It utilizes a tree of shapes to represent the image content. This representation provides a hierarchical tree for the objects contained in the level sets of the image. The hierarchical structure is utilized to select the candidate objects from the image. The Level set method provides a faster way to construct a tree of shapes. The boundaries of the selected objects are then compared with those of objects selected from prior images.

Keywords: level set methods, bias correction

I. INTRODUCTION
Segmentation plays a major role in digital image processing. Image segmentation algorithms are designed to segment an image into several regions so that the contents of each region represent meaningful objects. The segmentation results can then be utilized for post processing stages such as object recognition. Image segmentation simplifies post processing stages by focusing attention on each individual segment. Edge detection methods such as the Canny detector [1] were widely applied for this task. Edge detection methods utilize intensity gradients to detect the boundaries of objects. However, edge detection methods usually generate edges that are not closed contours, and this causes difficulties for later processing such as object recognition. Curve evolution methods [2][6] have been popular for image segmentation since the early 1990s. These methods evolve the initialized curve(s) to the boundaries of objects in an image. The above methods tend to segment the whole image into several regions, which is challenging for images with cluttered background. On the other hand, this is not always necessary in real applications. The user may be interested in finding only the location of objects of interest. For example consider a car moving on a road then the user can preferentially segment the car image by extracting the Volume 2, Issue 3 May June 2013

background. It is the main idea of preferential image segmentation. This is the key idea of preferential image segmentation, which means to preferentially segment objects of interests from an image and ignore the remaining portions of the image for this application. The idea of preferential image segmentation bears some similarities to object detection methods from images [3]. These methods detect the existence and rough location of objects in an image, e.g., in using a sparse, part-based object representation and a learning method. respectively. The image parsing method utilizes probabilistic methods to unify segmentation, detection and recognition. These methods, however, tend to classify/parse an whole image instead of meaningful objects. Furthermore, these methods are usually computationally intense. This method is motivated by the utilization of prior information in curve evolution models.

II. BACKGROUND
Section II-A shows how an image is represented using a tree of shapes. It also shows how a tree structure is introduced for image representation. Section II-B describes the relationship between color and geometry in natural images.Section II-C introduces the techniques of planar curve matching,which can be utilized to compare the boundaries of different objects. A. Image Representation Using trees of shapes. The tree of shapes represents images based on the techniques of contrast-invariant mathematical morphologies [4][7]. This method is based on the theory of image representation using connected components of set of finite perimeters in the space of functions with weakly bounded variations (WBV). The representation of an image using a tree of shapes utilizes the inferior or the superior of a level line to represent an object, and takes the boundary of the inferior area as the shape of the object. Therefore, only closed shapes are generated. This representation also provides a tree structure to represent the spatial relationship for the objects in an image. They have several advantages. First, they represent regions instead of curves in an image, which provide a way to handle the contents inside the regions. Second, Page 48

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
they are invariant to the contrast changes in an image, which may be caused by the change of lighting Third, closed boundaries are acquired for each upper level set or lower level set, which can be utilized for shape matching of the regions. The nesting of level sets provides an inclusion tree for an image. The inclusion tree from the family of upper level sets and the tree from the family of lower level sets, however, can be different if the connected components are directly utilized. The shapes in the lower level are spatially included in the shapes in the next higher level. The tree of shapes, therefore, provides a natural way to represent the spatial relationships between the shapes in the image. It is straightforward to find upper level sets and lower level sets in an image by thresholding. A tree of shapes can be further constructed by the nesting of level sets. However, this method is computationally intense. The image segmentation may be considerably difficult for images with intensity in homogeneities due to the overlaps between the ranges of the intensities in the regions to segmented. This makes it impossible to identify these regions based on the pixel intensity In general, intensity in homogeneity has been a challenging difficulty in image segmentation In the level set method, contours or surfaces are represented as the zero level set of a higher dimensional function, usually called a level set function. With the level set representation, the image segmentation problem can be formulated and solved in a principled way based on well-established mathematical theories, including calculus of variations and partial differential equations (PDE). An advantage of the level set method is that numerical computations involving curves and surfaces can be performed on a fixed Cartesian grid without having to parameterize these objects. Moreover, the level set method is able to represent contours/surfaces with complex topology and change their topology in a natural way. Most of region-based models are based on the assumption of intensity homogeneity. A typical example is piecewise constant (PC) models In level set methods are proposed based on a general piecewise smooth (PS) formulation originally proposed by Mumford and Shah. These methods do not assume homogeneity of image intensities, and therefore are able to segment images with intensity in homogeneities. However, these methods are computationally too expensive and are quite sensitive to the initialization of the contour which greatly limits their utilities. Edge-based models use edge information for image segmentation. These models do not assume homogeneity of image intensities, and thus can be applied to images with intensity in homogeneities However, this type of methods are in general quite sensitive to the initial conditions and often suffer from serious boundary leakage problems in images with weak object boundaries. Volume 2, Issue 3 May June 2013 From a generally accepted model of images with intensity in homogeneities, we derive a local intensity clustering property, and therefore define a local clustering criterion function for the intensities in a neighborhood of each point. This local clustering criterion is integrated over the neighborhood center to define an energy functional, which is converted to a level set formulation. Minimization of this energy is achieved by an interleaved process of level set evolution and estimation of the bias field. As an important application, our method can be used for Segmentation and bias correction of magnetic resonance (MR) images.A review is done for two wellknown region-based models for image segmentation secondly an energy minimization framework for image segmentation and estimation of bias field, which is then converted to a level set formulation for energy minimization is done.

III. THE LEVEL SET METHODS

These methods started with the work of Osher and Sethian (Front Propagating with Curvature-dependent speed: Algorithms based on Hamilton-jacobi formulations, Journal of Computational Physics 79, 1249, 1988) who showed how deformation could be modeled on discrete grids using level sets. Level set techniques, also known as implicit active contours have been the subject of active research in the last few years. The idea behind active contours (deformable models) is that the user specifies an initial guess for the contour, which is then moved by image driven forces to the boundaries of the desired objects. In such models, two types of forces are considered - the internal forces, defined within the curve, are designed to keep the model smooth during the deformation process, while the external forces, which are computed from the underlying image data, are defined to move the model toward an object boundary or other desired features within the image. There are two forms of deformable models. In the parametric form, also referred to as snakes, an explicit parametric representation of the curve is used. In contrast, the implicit deformable models, (also called implicit active contours or level sets), are designed to handle topological changes naturally. In a Level Set approach, instead of following the interface/contour directly, the contour is built into a surface. That is, contour is embedded as the zero level set of a higher dimensional function called the level-set function [14]. The level-set function is then evolved under the control of a partial differential equation. The motion of the interface is matched with the zero level set of the level set function, and the resulting initial value partial differential equation for the evolution of the level set function resembles a Hamilton-Jacobi equation. In this setting, curvatures and normal may be easily evaluated, topological changes occur in a natural manner, and the technique extends Page 49

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
trivially to 3 dimensions. At any time, the evolving contour can be obtained by extracting the zero level-set from the output. The level set method evolves a contour (in two dimensions) or a surface (in three dimensions) implicitly by manipulating a higher dimensional function, called the level set function _(x,t). The Evolving contour or surface can be extracted from the zero level set (x,t) = f_(x,t) = 0g Advantage of using this method is that topological changes such as merging and splitting of the Contour or surface are catered for implicitly, as can be seen below in Figure 2.1. The level set Method, since its introduction by Osher and Sethian in [13], has seen widespread application in Image processing, computer graphics (surface reconstructions) and Physical simulation (particularly uid simulation).The evolution of the contour or surface is governed by a level set equation. The solution tended to by this partial difererential equation is computed iteratively by updating at each time interval.

FIG 1.3 LEVEL SET METHOD

IV. BIAS CORRECTION

Methods of bias correction can be categorized into two classes: prospective methods and retrospective methods. Prospective methods aim to avoid intensity inhomogeneities in the image acquisition process. These methods, while capable of correcting intensity inhomogeneity induced by the imaging device, are not able to remove subject-induced intensity inhomogeneity [12]. In contrast, retrospective methods only rely on the information in the acquired images. Therefore, they can also remove intensity inhomogeneities regardless of their sources. Early retrospective methods include those based on filtering, surface fitting, and histogram.Segmentation based methods are more attractive, as they unify segmentation and bias correction within a single framework These methods consist of the interaction between segmentation and bias correction, which benefits each other to yield final segmentation and bias correction.

A. Color and Geometry in Mathematical Morphologies An improved space HLS (IHLS) is utilized for the color model. The color in every pixel is represented with three channels (Y,S,H), which corresponds to the gray level, saturation and hue respectively. The IHLS space[10], compared to other spaces such as HLS and HSV, has the property of a well-behaved saturation coordinate. The IHLS space always has a small numerical value for near-achromatic colors, and is completely independent of the brightness function. For a pixel with color(R,G,B) in the(R,G,B) space, and the corresponding pixel with color(Y,S,H) , the transformation from the RGB space to the IHLS space is Y=0.2126R + 0.7152G + 0.0722B S=MAX(R,G,B)-MIN(R,G,B) H= { 360-H, if B>G H, OTHERWISE Where H= This method uses the transformation and the total order in to extract the shapes and build the tree of shapes for color images. The inverse transformation from the IHLS space to the RGB space is not utilized. The color image is synthesized so that its gray version contains no shape information. Using the IHLS color space and the total order a tree of shapes is built for the color image. B. Planar Curve Matching The method in defines the shape of a curve as a conjunction of shape elements and further defines the shape elements as any local, contrast invariant and affine invariant part of the curve. These definitions are oriented to provide invariance to noise, affine distortion, contrast changes, occlusion, and background. The shape matching between two images are designed as the following steps. 1) Extraction of the level lines for each image. The level set representations are utilized here for the extraction. The level lines are defined as the boundaries of the connected components as shown before. 2) Affine filtering [11] of the extracted level lines at several scales. This step is applied to smooth the curves

FIG 1.1 LEVEL SET METHOD

FIG 1.2 LEVEL SET METHOD

Volume 2, Issue 3 May June 2013

Page 50

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 3, May June 2013 ISSN 2278-6856
using affine curvature deformation to reduce the effects of noise. 3) Local encoding of pieces of level lines after affine normalization. Both local encoding and affine normalization are designed for local shape recognition methods. This step will help to deal with occlusions in real applications. 4) Comparison of the vectors of features of the images. Euclidean distance is utilized to compare the feature vectors. The performance of curve matching between two curves is calculated after affine filtering, curve normalization and local encoding. Trans. Image Process., vol. 9, no. 5, pp. 860872, May 2000 [8] Y.-G. Wang, J. Yang, and Y.-C. Chang, Colortexture image segmentation by integrating directional operators into jseg method, Pattern Recogn. Lett., vol. 27, no. 16, pp. 1983 1990, 2006. [9] S. Y. Yu, Y. Zhang, Y. G. Wang, and J. Yang, Unsupervised color-texture image segmentation, Journal of Shanghai Jiaotong University (Science), vol. 13, no. 1, pp. 7175, February 2008. [10] A. Hanbury and J. Serra, A 3d-Polar Coordinate Colour Representation Suitable for Image Analysis, PRIP, T.U.Wien, Tech. Rep. PRIP-TR-77, 2002. [11] J. L. Lisani, L. Moisan, P. Monasse, and J. M. Morel, On the theory of planar shape, Multiscale Model Simul., vol. 1, no. 1, pp. 124, 2003. [12] C. Li, R. Huang, Z. Ding, C. Gatenby, D. Metaxas, and J. Gore, A variational level set approach to segmentation and bias correction of medical images with intensity inhomogeneity, in Proc. Med. Image Comput. Comput. Aided Intervention, 2008, vol. LNCS 5242, pp. 10831091, Part II. [13] L. Evans, Partial Differential Equations. Providence, RI: Amer. Math. Soc., 1998. [14] A Level Set Method for Image Segmentation in the Presence of Intensity Inhomogeneities With Application to MRI Chunming Li, Rui Huang, Zhaohua Ding, J. Chris Gatenby, Dimitris N. Metaxas, Member, IEEE, and John C. Gore

V. SUMMARY AND FUTUREWORK

The method utilizes both the intensity and shape prior information by means of the tree of shapes and boundary matching. It is invariant to contrast change and similarity transformations such as scale, rotation and translation. Future research on segmentation methods will be focused on the multiscale analysis is also applied to large image databases to examine its performance. An advantage of the level set method is that numerical computations involving curves and surfaces can be performed on a fixed Cartesian grid without having to parameterize these objects. A systematic evaluation of the proposed method will be performed on large image databases, and the results will be provided

REFERENCES
[1] J. F. Canny, A computational approach to edge detection, IEEETrans. Pattern Anal. Mach. Intell., vol. 8, no. 6, pp. 679698, Jun. 1986. [2] J. Sethian, Level Set Methods and Fast Marching Methods, 2nd ed. Cambridge, U.K.: Cambridge Univ. Press, 1999, Cambridge Monograph.on Applied and Computational Mathematics [3] S. Agarwal, A. Awan, and D. Roth, Learning to detect objects in images via a sparse, part-based representation, IEEE Trans. Pattern AnaysisMach. Intell., vol. 26, no. 11, pp. 14751490, Nov. 2004. [4] P. Monasse, Morphological representation of digital images and application to registration, Ph.D. dissertation, Ceremade, Paris, 2000. [5] Yining Deng and B.S. Manjunath, Unsupervised Segmentation of Color-Texture Regions in Images and Video, IEEE Trans on Pattern Analysis and Machine Intelligence, vol. 22, no. 6, p.939-954, 2001 [6] D. Cremers and G. Funka-Lea, Dynamical statistical shape priors for level set based sequence segmentation, Variational and Level Set Methods Comput. Vis., pp. 210221, 2005. [7] P. Monasse and F. Guichard, Fast computation of a contrast-invariant image representation, IEEE

Author
Mr. Sasi Kumar has completed his Master of computer applications in MS University, Thirunelveli. Currently working as technical lead in Verizon Data Services Pvt Ltd, Managing, Mentoring and Leading a team of 15 members. Have given internal technical trainings in various technologies to various teams in the organization. His area of interests are Image Processing, Computer Networks .Have trained 200 staffs of central bank of India,Thiruvananthapuram in computer subjects. And also have written books in Java, J2EE, C,C++,Data Structures, DBMS, Computer Networks for C-DAC government of kerala, Thiruvananthapuram.

Volume 2, Issue 3 May June 2013

Page 51