{"title":"Real-time 3D video synthesis from binocular capture system based on commodity graphic hardware","authors":"Zhihong Liao, Xudong Xie, Qionghai Dai","doi":"10.1109/3DTV.2009.5069620","DOIUrl":null,"url":null,"abstract":"In this paper, a real-time 3D video synthesis method suitable for implementation on commodity graphic hardware is presented. The system consists of pre-calibrated binocular stereo cameras and an NVIDIA GeForce 8 Series graphic card. Recently, most research has focused on improving the quality of depth maps, which is usually time-consuming and unsuitable for real-time reconstruction. In our method, we combine a plane-sweeping algorithm with view synthesis, and thus the new views between the captured images can be real-time generated without generating the intermediate depth map. The Programmable Fragment Shader technology is used to accelerate the process of Sum-of-Square-Difference (SSD) dissimilarity measure of the correlation pixels and the final color calculation. Experimental results show that our method can efficiently implement the real-time 3D view synthesis.","PeriodicalId":230128,"journal":{"name":"2009 3DTV Conference: The True Vision - Capture, Transmission and Display of 3D Video","volume":"13 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2009-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2009 3DTV Conference: The True Vision - Capture, Transmission and Display of 3D Video","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/3DTV.2009.5069620","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
In this paper, a real-time 3D video synthesis method suitable for implementation on commodity graphic hardware is presented. The system consists of pre-calibrated binocular stereo cameras and an NVIDIA GeForce 8 Series graphic card. Recently, most research has focused on improving the quality of depth maps, which is usually time-consuming and unsuitable for real-time reconstruction. In our method, we combine a plane-sweeping algorithm with view synthesis, and thus the new views between the captured images can be real-time generated without generating the intermediate depth map. The Programmable Fragment Shader technology is used to accelerate the process of Sum-of-Square-Difference (SSD) dissimilarity measure of the correlation pixels and the final color calculation. Experimental results show that our method can efficiently implement the real-time 3D view synthesis.