Motion detail preserving optical flow estimation

Abstract

We discuss the cause of a severe optical flow estimation

Occlusion-aware temporal frame interpolation in a highly scalable video coding setting

“…To show the applicability of the proposed scheme on natural sequences, we need to estimate motion fields that satisfy our requirements. We found that Xu et al 's [14] motion detail preserving (MDP) optical flow algorithm provides motion fields of sufficient quality to work with our proposed framework; the parent motion field M a→c is estimated using the default parameters of their implementation.…”

“…Example results for estimated motion on a natural sequence with reasonably complex motion. Panel (a) shows the motion fieldM a→c , estimated using [14] with the default parameters; Panel (b) shows the breakpoint field (at the second coarsest spatial level for visualization), which was estimated onM a→c using the breakpoint estimation method described in [16]. Panel (c) shows the union of the estimated disocclusion masks, where yellow and cyan indicate that the pixel is not visible in the previous ( f a ) and future ( f c ) frame, respectively.…”

“…Recently, there have been various proposals from the computer vision community on how such motion fields can be estimated. Xu et al [14] propose a motion detail preserving optical flow algorithm (MDP), which encourages sharp discontinuities in the motion field. Wulff and Black [15] propose a layered motion model, which is able to obtain piecewise-smooth motion fields with sharp discontinuities on sequences that are heavily affected by motion blur.…”

Two Algorithms for Motion Estimation from Alternate Exposure Images

“…This approach is refined by Sand and Teller [7] in the context of particle video. Xu et al consider the in-flow into a traget pixel as occlusion measure [9] (see also [17]). While explicit occlusion handling is incorporated, the moment of occlusion cannot be determined.…”

“…A scene point and its motion contribute to a motion-blurred image exactly for as long as the point is not occluded. Only recently have optical flow algorithms begun to address occlusion [6][7][8][9], assigning occlusion labels per pixels. The moment of occlusion, however, cannot be easily determined from short-exposure images.…”