Invariant spatial chromatic processors for region image description

Abstract

A new region image descriptor is proposed. This descriptor arises from the chromatic methodology to extract the features from the images by applying new type of processors called 'Invariant spatial chromatic processors'. The properties of the spatial chromatic processors and the proposed approach of making them invariant are presented. A database of more than one thousand gray scale images is used to test the proposed descriptor with rotation, translation, scaling, and noise effects, where the Moment Invariant is also used to benchmark the results. The simulation results show, a comparable performance between the proposed descriptor and the Moments Invariant during affine effects with better performance during noise effect.