STRUCTURAL MATCHING OF IMAGES BY HASHING GRADIENT SINGULARITY DESCRIPTORS

摘要

STRUCTURAL MATCHING OF IMAGES BY HASHING GRADIENT SINGULARITY DESCRIPTORS The method for matching digital images of the same article in a data processing unit comprises the steps of transforming each digital image of an article into a topographic map of local divergence of the luminance gradient vector field, of detection of singularities or extrema of local divergence of the luminance gradient vector field, these singularities corresponding to points of interest in said digital image, and for each point of interest detected, encoding the values of the singularity of the gradient field located on several concentric rings centered on the point of interest to derive a vector of digital data (210) and transforming said vector into a digital hash key (220) by means a family of LSH Cosine hash functions. Figure to be published: Fig. 2A