Font Size: a A A

Depth Extraction And Virtual View Synthesis From Multi-view Stereo Images

Posted on:2011-02-02Degree:MasterType:Thesis
Country:ChinaCandidate:J ShiFull Text:PDF
GTID:2178330338483663Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
With the development of related disciplines, stereoscopic video becomes a new direction in the field of image and video processing. Naturally the stereoscopic television technology becomes the fourth generation TV technology. Stereoscopic television technology includes acquisition, processing, transmission, displaying and watching of stereoscopic images, which is totally five parts. This paper first introduces the imaging principle, generation technology and displaying technology of stereoscopic images, and explores the development process of stereoscopic television from domestic to abroad and looks forward to their development prospects. Then, it makes a complete method process, which can be used in stereoscopic images generation and displaying, and researches deeply around the key issues of depth map estimation and viewpoint generation, obtained the following results:1. Against the depth information extraction part of the process, this paper proposes two algorithms of depth map extraction based on the windows and improved window-based cross searching. This paper analyzes and resolves the block choice in regional matching, and combines local matching algorithm and global matching idea to achieve a better effect of the depth map.2. Against virtual viewpoint generation part, this paper proposes algorithm based on depth map, which interacts with depth map extraction algorithm, and then uses parallax calibration to resolve the depth-jumping problem in naked eye viewing stereoscopic images. Not only wearing glasses viewing but also naked eye viewing in watching the stereoscopic images can achieve obvious stereoscopic effect, clarity image and comfort viewing.3. This paper applies the above-mentioned algorithms into two-viewpoint video to extract depth video. Moreover, it successfully resolves the problem of time-domain frames flicker in depth video. Through the above steps, it completes each step of process.This paper does a lot of experiments to realize the depth information extraction o both multi-view images and multi-view video, in order to achieve virtual viewpoints generation based on depth information. Finally, this paper can complete stereoscopic images displaying based on virtual viewpoints. This paper can meet the stereoscopic images generation and displaying method based on the depth map.
Keywords/Search Tags:disparity, depth map, stereo matching, virtual viewpoint, parallax calibration, time-domain smoothing
PDF Full Text Request
Related items