Shape Matching and Object Recognition Using Dissimilarity Measures with Hungarian Algorithm

The shape of an object is very important in object recognition. Shape matching is a challenging problem, especially when articulation and deformation of a part occur. These variations may be insignificant for human recognition but often cause a matching algorithm to give results that are inconsistent with our perception. In this paper, we propose an approach to measure similarity between shapes using dissimilarity measures with Hungarian algorithm. In our framework, the measurement of similarity is preceded by (1) forming the shapes from the images using canny edge detection (2) finding correspondence between shapes of the two images using Euclidean distance and cost matrix (3) reducing the cost by using bipartite graph matching with Hungarian algorithm. Corresponding points on two dissimilar shapes will have similar distance, enabling us to solve an optimal assignment problem using the correspondence points. Given the point correspondence, we estimate the transformation that best aligns the two shapes; regularized thin plate splines provide a flexible class of transformation maps for this purpose. The dissimilarity between the two shapes is computed as a sum of matching error between corresponding points, together with a term measuring the magnitude of the aligning transform. By using this matching error, we can classify different objects. Results are presented and compared with existing methods using MATLAB for MNIST hand written digits and MPEG7 images.