000149301 001__ 149301
000149301 005__ 20190416220650.0
000149301 037__ $$aREP_WORK
000149301 245__ $$aAn optimal first-order solver for the TV-$L_1$ optical flow problem
000149301 269__ $$a2010
000149301 260__ $$c2010
000149301 336__ $$aReports
000149301 500__ $$aWe are currently improving our GPU implementation. Hence, GPU source code will be made available in a near future.
000149301 520__ $$aIn this study, we address the problem of computing efficiently a dense optical flow between two images under a total variation (TV) regularization and an $L_1$ norm data fidelity constraint using a variational method. We build upon Nesterov's framework for convex minimization. By keeping in memory the solution estimated at the previous iteration, this framework yields convergence rates of one order of magnitude faster than existing algorithms, hence is computationally more efficient. We show how to adapt this method to the TV-$L_1$ problem by using a smoothed reformulation of the TV norm to make it continuously differentiable. This relaxation is controlled by a single parameter whose effects are also studied in this paper. Finally, we demonstrate how this fast algorithm can be easily implemented on modern graphics hardware (GPU) using the recently proposed OpenCL Application Programming Interface (API) in order to achieve further speedups.
000149301 6531_ $$aoptical flow
000149301 6531_ $$aTotal Variation (TV)
000149301 6531_ $$aconvex optimization
000149301 6531_ $$agpu
000149301 6531_ $$aOpenCL
000149301 6531_ $$aLTS2
000149301 700__ $$0242926$$g185083$$aD'Angelo, Emmanuel
000149301 700__ $$0242927$$g179918$$aPuy, Gilles
000149301 700__ $$aVandergheynst, Pierre$$g120906$$0240428
000149301 8564_ $$uhttp://lts2www.epfl.ch/~dangelo/opticalflow.html/$$zURL
000149301 8564_ $$uhttps://infoscience.epfl.ch/record/149301/files/FistaFlow.pdf$$zn/a$$s2904598$$yn/a
000149301 909C0 $$xU10380$$0252392$$pLTS2
000149301 909CO $$ooai:infoscience.tind.io:149301$$qGLOBAL_SET$$pSTI$$preport
000149301 917Z8 $$x185083
000149301 917Z8 $$x185083
000149301 937__ $$aEPFL-REPORT-149301
000149301 973__ $$sPUBLISHED$$aEPFL
000149301 980__ $$aREPORT