Vision significance detection method based on Weber's law and center-periphery hypothesis

Abstract

The invention relates to a vision significance detection method based on Weber's law and center-periphery hypothesis. The conventional method is low in resolution, incomplete in extracted object outline and high in computational complexity. The method comprises the following steps of: extracting color component graphs of original images in a CIELAB space by adopting a color transformation method; calculating horizontal gradient difference excitation values and vertical gradient difference excitation values of pixel points in the 1 color component graph, the color component graph and the b color component graph according to the Weber's law; calculating the difference excitation value of a random gradient direction according to the horizontal gradient difference excitation values and the vertical gradient difference excitation values, and counting a difference excitation value histogram; and finally, establishing local significance excitation vectors of the pixel points to obtain local significance judgment values and overall significance excitation values, and calculating the significance judgment values according to the local significance judgment values and the overall significance excitation values. By the method, vision significance graphs with the same resolution as the input images can be acquired, and stronger response in a significance region can be realized.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a bottom-up visual saliency detection method based on Weber's theorem and central-peripheral hypothesis. Background technique [0002] Visual saliency detection is a key step in the visual selective attention mechanism, so that the computer only needs to perform relevant calculations and processing on the visual saliency area in the image, thus providing a fast and effective method for reducing computational complexity. It has been widely used in machine vision, image processing, intelligent robots and other fields. [0003] The bottom-up (spatial) visual saliency calculation model was first proposed by L.Itti et al. Based on the feature integration theory and the central-peripheral assumption, the model decomposes the input image into multiple independent feature spaces such as brightness, color, and direction, and extracts the saliency map of each feature space, and ...

AI Technical Summary

Problems solved by technology

Unlike biologically inspired saliency models, these are often purely computational models, inspired by the biological concept of central-peripheral differences, but not based on any biological model

Images