A Polar-Fourier-Wavelet Transform for Effective CBIR

Problems of effective rapid and reliable CBIR (Content Based Image Retrieval) are discussed in the paper, generally concerning the so called “primary” CBIR, i.e. the image query by example into a large database of images of hallmark type. A novel Polar-Fourier-Wavelet transform (PFWT) is proposed herein, allowing both high invariance towards accidental input image rotation and low correlation among the transform coefficients, the latter expected significantly to reduce the retrieval errors. Beside the theoretical aspects of the PFWT, result analysis is reported on tests over three IDBs from the real practice.