Learning wavelet networks based on Multiresolution analysis: Application to images copy detection

This paper proposes a new method for images contentbased copy detection through a wavelets network based on multiresolution. The system is not only able to detect image copies, but it is also able to detect similar images to the original reference. The idea of this method is to project both the original images and the test one on a MRWN (MultiResolution Wavelets Network). In fact, this kind of wavelets network is a hybrid one constituted by a combination of a neural network and a wavelet network. This later is used to measure intelligently the similarity through two filters: approximation filter and details filter. As the matter of fact, the use of wavelet networks has allowed us to create an efficient and robust system for image copy detection.