0722 Progress Report B 00901099 NTUEE 3 rd

07/22 Progress Report B 00901099 NTUEE 3 rd Hsiao Yi

Implementation of a DVC scheme Objective: want to know how much performance we can get if we use the greatest motion estimation methods, such as: optical flow, global motion estimation, affine motion estimation, motion estimation in HEVC (variable block size, 1/4 precision) Construct a DVC scheme based multiple side information generation including optical flow, noise model learning [1] X. Huang , L. L. Raket , H. V. Luong , M. Niesen , F. Lauze and S. Forchhammer "Multi-hypothesis transform domain wyner-ziv video coding including optical flow", Proc. IEEE Int. Workshop Multimedia Signal Process. , pp. 1 -6 2011 [2] H. Luong, L. L. Raket, X. Huang; S. Forchhammer, "Side Information and Noise Learning for Distributed Video Coding Using Optical Flow and Clustering, " IEEE Transactions on Image Processing, vol. 21, no. 12, pp. 4782, 4796, Dec. 2012 Substitute or add motion estimation methods Mainly use Open. CV -> contain optical flow and machine-learning features

Optical Flow Implementation Have implemented some optical flow algorithms (Lucas Kanade method with pyramid refinement、Gunnar Farneback's algorithm、Dual TV L 1、Simple Flow(still have bugs)) Detail of implementation : ->Lucas Kanade method with pyramid refinement--> Matlab (Matlab File Exchange) ->Gunnar Farneback‘s algorithm、Dual TV L 1、Simple Flow --> C++(Open. CV) Simple Flow: Michael W. Tao, Jiamin Bai, Pushmeet Kohli, and Sylvain Paris. "Simple. Flow: A Non-iterative, Sublinear Optical Flow Algorithm". Computer Graphics Forum (Eurographics 2012), 31(2), May 2012.

Result present Visualized by using http: //vision. middlebury. edu/ evaluation code(munsell color system) Gunnar Farneback Dual TV L 1 Simple Flow Lucas-Kanade