Accurate Optical Flow via Direct Cost Volume Processing

Abstract

We present an optical flow estimation approach that operates on the full four-dimensional cost volume. This direct approach shares the structural benefits of leading stereo matching pipelines, which are known to yield high accuracy. To this day, such approaches have been considered impractical due to the size of the cost volume. We show that the full four-dimensional cost volume can be constructed in a fraction of a second due to its regularity. We then exploit this regularity further by adapting semi-global matching to the four-dimensional setting. This yields a pipeline that achieves significantly higher accuracy than state-of-the-art optical flow methods while being faster than most. Our approach outperforms all published general-purpose optical flow methods on both Sintel and KITTI 2015 benchmarks.

Published at the Conference on Computer Vision and Pattern Recognition (CVPR 2017)

Cited by in corpus (11)

• Deep Stereo Matching with Explicit Cost Aggregation Sub-Architecture
• Occlusion Aware Unsupervised Learning of Optical Flow
• SelFlow: Self-Supervised Learning of Optical Flow
• Automatic Generation of Dense Non-rigid Optical Flow
• Deep Motion Boundary Detection
• Flow2Stereo: Effective Self-Supervised Learning of Optical Flow and Stereo Matching
• KFNet: Learning Temporal Camera Relocalization using Kalman Filtering
• FusionNet and AugmentedFlowNet: Selective Proxy Ground Truth for Training on Unlabeled Images
• Learning Optical Flow from a Few Matches
• Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation
• Optical Flow Super-Resolution Based on Image Guidence Using Convolutional Neural Network