A Region Based Image Matching Method with Spatial Constraints

In this paper, a new region-based image matching method was proposed to find the user defined regions in other images. Not only local features and overall features are used, but spatial constraints are also employed. Experimental data were composed of 1000 images from Corel and 1000 I frames extracted from CCTV and MPEG-7 videos. Three matching strategies were compared in experiments. They were overall feature only, color and spatial constraints, and color/texture with constraints. As well, lots of noise was introduced to test the robustness of the proposed method. Experimental results showed that the method was feasible and effective.