An image retrieval system based on adaptive wavelet lifting

We propose an algorithm for content-based image retrieval of grayscale images of objects against a background of texture. It employs feature vectors based on moment invariants of detail coefficients produced by the lifting scheme. The prediction filters in this scheme are chosen adaptively: low order (small stencils) near edges and high order elsewhere. The aim is an algorithm that retrieves similar images of an object irrespective of translation, rotation, reflection or re-sizing of the object, lighting conditions and the background texture. We discuss and propose both linear and nonlinear transforms of the feature vectors in order to develop a suitable metric for measuring the distance between them. We present preliminary results with respect to an artificial database.