Matching Objects between Multiple Cameras Based on Wavelet Salient Features

This paper proposes a novel and effective method to match objects between multiple cameras, which is based on wavelet salient features consisting of color information and salient values at salient points. Learning templates are set up and updated based on salient features in one camera, and used to match objects in the other camera without any geometric constraints. Experiments show the effectiveness of the proposed method in matching different views of the same objects, even when occlusion occurs.