Method for the characterization of a digital image and the applications thereof

摘要

A method of characterizing a digital image, in which a point of interest (PI) in the detected in the image. For each θ of a plurality I of different prescribed orientations in the digital image a family {f_θ^(k)(n)}_1≦K≦K of K one-dimensional lines f_θ^(k)(n) of points is extracted from a region (RPI) of the image along K straight lines of orientation (θ), and for each line f_θ^(k)(n) and for each sub-sampling level s, a foveal energy signal E_θ,s^(k)(n) is calculated using foveal wavelets. The signals E_θ,s^(k)(n)) on the K lines and their arguments n are used to calculate a foveal signature SI characterizing the working region.