Shape Matching by Region Grouping

In this paper, we propose a novel region grouping approach to shape matching. It is proposed as an alternative region based approach to the traditional edge based shape matching using distance transforms. It has the advantage of obtaining a higher detection rate and obtaining meaningful object segmentation simultaneously. Each image is first segmented into image regions, and possible matches are found among the regions based on a proposed probabilistic similarity measure to the exemplars. We evaluate this new approach against edge based matching using Chamfer and Hausdorff distances. Pros and cons of the proposed method are discussed.