Image Recognition And Matching Based On Weber Local Feature And Shape Context

In computer vision, people regard descriptor studies as the essential and fundamental parts of the research. Along with the broadly applications of machine learning and computer vision in various fields; image feature descriptor has been widely recognized to many technicians in breadth and depth. The feature descriptor's character and robustness are directly connected with the experimental performance of recognition ability and detection speed.Chen, et al. in 2010 proposed a Weber Law based local descriptor. It is based on the fact that human perception of a pattern depends not only on the change of a stimulus but also on the original intensity of the stimulus. It is a simple, powerful and robust local descriptor, named the Weber Local Descriptor (WLD). Specifically, WLD consists of two main components: differential excitation and gradient orientation. The differential excitation component is a function of the ratio between the relative intensity differences of a current pixel against its neighbors, and the intensity of the current pixel. The orientation is the gradient of the current pixel. Experimental results on the CBCL and MIT-CMU face databases show a WLD impressively efficiency and robustness performance. While, classic WLD differential excitation lacks the explicit texture and edge detail discrimination of images with the hue and saturation that vary smoothly. The improved WLD differential excitation contains two main parts: 1) original image based frequency domain energy entropy, 2) Gaussian band pass filter, which can enhance the robustness and the discrimination of the WLD differential excitation.Another main issue of this paper is shape context based shape matching algorithm. By adopting shape matching by shape context algorithm, which is polar coordinates histogram-feature, the performance of which is taken into consideration. Herein, by applying polar coordinates based shape matching by shape context, the recognition and matching of such can achieve a pleasant result in shape matching and face recognition. In shape matching by shape context algorithm measurement of similarity is preceded by two procedures: 1) searching for correspondences between points on the two shapes, 2) adopting the correspondences to estimate a transform function. For the purpose of solving the correspondence problem, the shape context in each point is needed The shape context at a reference point captures the distribution of the remaining points relative to it, hence, enabling the algorithm to solve for correspondences as an optimal assignment problem.As state previously, by realizing the WLD face recognition algorithm and shape matching by shape context algorithm, the improvement of WLD and shape matching algorithm become feasible. Also, the performance of improved WLD differential excitation has shown a better result than the classic WLD differential excitation based face recognition.