Segmented tree disparity map acquisition method based on improved matching cost

Abstract

The invention discloses a segmented tree disparity map acquisition method based on improved matching cost. The method comprises the following steps: carrying out pixel point acquisition on a left image and a right image; establishing a matrix; performing cost calculation; converting the undirected connected graph; converting the minimum spanning tree; calculating the weight between adjacent pixelpoints; cost aggregation; acquiring left and right disparity maps wherein left and right pixels are consistent; introducing depth information and color information; performing cost aggregation again;obtaining a final disparity map. For the defect that ambiguity matching is easily generated in a weak texture region, color and gradient multi-dimensional information components are introduced, so that the weight distribution of a window is more reasonable, and the cost amount of pixel points in the weak texture region is enhanced; a novel grouping strategy is provided, and pixel points conformingto parallax consistency are divided to the same spanning tree; and a more accurate matching result can be generated while the time efficiency is kept.

Description

technical field

[0001] The invention relates to the technical field of image processing, in particular to a method for acquiring a segmented tree disparity map based on an improved matching cost. Background technique

[0002] In the process of exploring the external world, human beings obtain information in various ways. However, at least 80% of the information comes from the visual part. As an important visual organ, the eyes can transmit the captured scenes to the brain for analysis and processing. With the continuous development of science and technology and the maturity of related theories, researchers hope that computers can also realize the functions of the human visual system. Not only can images of real scenes be obtained, but at the same time, it is hoped to be able to identify image information and analyze and process important information, thus the discipline of computer vision was born. As a popular research direction in this discipline, stereo vision technolo...

Images