Article Preview
TopIntroduction
Computer vision is a field in which machines are trained and equipped with the ability to depicting human vision in the machines that is to make the machines view and recognize objects in a scene. In the recent year computer vision have made enormous progress in this field to achieve high quality visual perception and object recognition. To recognize an object, there are several properties that can be used for the purpose of recognition and classification, like object shape, object color, object texture and object brightness. Of all these properties shapes is the most intrinsic feature used for recognition of objects. Shape representation is done using two major approaches, one the boundary based approach which uses contour information and the second approach needs a holistic representation, requiring general information about the shape (Daliri.R and Vincent.T, 2007, p. 1782).
Helin.D, Bulent.S and Yucel. Y (2010), proposed a method using subspace approach, which solves the ambiguity of pose normalization with continuous PCA coupled with the use of feasible axis labeling and reflection (p.865). Bribiesca. E and Wilson.R.G (1997), presented an approach for 2D shape object dissimilarity. The shape of the different objects to be compared is mapped to a representation invariant under translation, rotation and Scaling (p. 107). Bandera et al. (1999), proposed an algorithm, where contours are represented by their curvature function, decomposed in the Fourier domain as linear combination of a set of representative object and object are identified by multilevel clustering (p. 49). Kumar and Rockett (1997), proposed a method representing scaling, translation and rotation based on the invariance properties of angle of the triangle, which are used to construct signature histogram of local shape (p. 235). Guerra. C (1998), presented an approach using reconfigurable mesh architecture with horizontal and vertical broadcasting. The object models are described in terms of a convex/concave multi scale boundary decomposition that is represented by a tree structure (p. 83). Khalil and Bayoumi (2000), proposed a method to recognize 2D object under translation, rotation and scale transformation, using the technique based on the continuous wavelet transform and neural networks (p. 863). Mcneill and Vijaykumar (2005), present a corresponding-based technique for efficient shape classification and retrieval. Shapes are represented by a large number of points on the boundary, the points lie at fixed intervals on the boundary or radial angle, which gives a robust description of shapes (p. 1483).