Stereo vision saliency calculation method based on left and right monocular receptive fields and binocular fusion

Abstract

The invention relates to a stereoscopic vision significance calculating method based on left and right monocular receptive field and binocular fusion and belongs to the technical field of computer vision. Through simulating the processing process of a human stereoscopic vision system, firstly, left and right monocular receptive field models are respectively built, corresponding characteristic patterns are obtained, then, left and right eye characteristic patterns with the same characteristics and result patterns with different characteristics are fused, in addition, the pixel significance is regulated on the basis of center preference and prospect preference, and finally, a significance distribution pattern most closely approaching to the human stereoscopic vision attention is obtained. The stereoscopic vision significance calculating method has the innovation points that the influence of two common phenomena of the center preference and the prospect preference on the human stereoscopic vision attention is added, so the whole method can perfectly conform to the processing process of the human vision system. The stereoscopic vision significance calculating method provided by the invention can perfectly simulate the processing process of the human stereoscopic vision system, and the interesting region in stereoscopic images can be more effectively extracted.

Description

technical field [0001] The invention relates to a stereoscopic saliency calculation method based on left and right monocular receptive fields and binocular fusion, and belongs to the technical field of computer vision. Background technique [0002] In the vast world faced by human beings, there are massive amounts of information around human beings at any time. However, on the one hand, the human sensory system has limited capacity and cannot receive all information at the same time; on the other hand, compared with the sensory system, the brain's information processing capacity is very limited, and there is a bottleneck problem between the two. Visual attention is an essential part of the mental system in order to effectively receive and process information. [0003] When facing a complex scene, the human visual system (HVS) can quickly selectively filter out non-important information in the visual scene, so as to focus on a few visually salient objects. This process is ca...

AI Technical Summary

Problems solved by technology

[0011] However, most stereoscopic saliency calculation methods only extract the saliency map of the stereoscopic image on the left eye image, and do not consider the binocular fusion problem, which is not completely in line with the human stereoscopic vision system processing process.

On the other hand, most of the stereo vision saliency calculation methods use images taken by the author himself, without corresponding manual annotation results, and there is no public stereo image evaluation data set, so quantitative comparison cannot be done

[0012] In general, most of the bottom-up stereo vision saliency calculation methods are extended based on the two-dimensional visual saliency calculation method, which simulates the monocular receptive field model well, and conforms to part of the human visual processing process, but most of the calculation methods It only takes depth as a feature, does not consider the problem of binocular fusion, and does not include the important process of binocular interaction in the human stereoscopic vision system

Images