詹宝珠教授报告者.PPT

Electrical Engineering Department of National Cheng Kung University * Smart Media Intelligent Living Excellence Lab. Electrical Engineering Department of National Cheng Kung University Smart Media Intelligent Living Excellence Lab. * * 2012/08/14 指導教授：詹寶珠教授 報告者：陳正旻 期末DEMO line segment的GPU實現 * Outline 演算法流程介紹(附實作結果) 時間比較和加速倍數 實作影片 總結 Flow chart * Algorithm Line segment construction Two rules Color difference : Spatial distance : * (TIME)Line segment construction * Flow chart * Algorithm Initial Disparity Computation Matching cost : CAD : CCENSUS : Hamming distance of the two strings that stand for p and pd * String R. Zabih and J. Woodfill, “Non-parametric local transforms for computing visual correspondence,” in Proc. ECCV, 1994, pp. 151–158. d=1 Left Right 2 (TIME)Initial Disparity Computation Matching cost * Algorithm Cost aggregation Winner-take-all * d=1 Left Right (TIME)Cost aggregation Winner-take-all * 初步結果 左圖的Depth 右圖的Depth * Flow chart * Algorithm Seed Pixel Detection Two rules * (TIME)Seed Pixel Detection * Flow chart * Algorithm Scanline Propagation Aa , if p is in occluded regions. d Ss , if p is near depth discontinuities. a is replaced by the linearly interpolated results of and in all other conditions. * (TIME)Scanline Propagation * Scanline Propagation之後的結果 * Flow chart * Algorithm Disparity Refinement Vertical voting : We collect ?? disparity votes in a vertical line segment that starts at pixel and ends at pixel ,if the color difference is smaller than the threshold value ?? , the votes on disparity increase by 1. c * (TIME)Vertical voting * Vertical voting之後的結果 * 再加入BilateralFilter之後的最終結果 * GPU運算時間 CPU運算時間 * 時間比較和加速倍數 Line Segment 漢明距離 單點cost 總cost CPU 0.114000