Research On Passive Acquisition Of Depth Map

Depth acquisition of stereo image is one of the most important problems in the field of stereo vision,which can be divided into two part of method : active depth acquisition and passive depth acquisition.The active depth acquisition method refers to the method of obtaining depth information by using TOF camera(camera Time-of-Flight)or other depth image acquisition devices to directly capture the scene depth map.Passive depth acquisition method mainly includes stereo matching algorithm.By matching the corresponding pixels in the images obtained from different viewpoints,the corresponding disparity value is obtained,which can reduce the depth information of the whole scene.This paper is research on stereo matching algorithm of passive depth acquisition method.The main work of this paper is as follows :First we proposed an improved local stereo matching algorithm.After that,we proposed an optimization method of stereo depth map based on image segmentation.The main work of this paper is as follows:1.In view of the limitations of the traditional local stereo matching algorithm in image occlusion,depth discontinuities,weak texture and repetitive texture region.An improved local stereo matching algorithm based on similarity measure function is proposed.First,a new similarity measure function is constructed,which is based on the combined of the information of color gradient and spatial distance between pixels and a non parametric transformation method.In the cost aggregation process,the initial disparity map is obtained by using the adaptive window based on the cross frame and guide image filtering method;2.Using mean shift algorithm for the image segmentation of initial disparity map.The false matching points in the disparity map are eliminated by the method of surface interpolation,which consider the information of segmentation result and confidence map,and then get the final result.The test result have evaluated on the Middlebury platform.The average false matching rate of the occlusion region and weak texture region is reduced by about 40% compared with the classical algorithm.In addition,the complexity of the algorithm has been significantly reduced,the average computation time is 28.2s.The final results show that proposed algorithm has a better result on the disparity information of the occlusion region and the depth discontinuity region.