We propose the Accelerated Cost Volume Filtering (ACF) method, which speeds up the traditional Cost Volume Filtering (CF) method. ACf identifies salient subvolumes in the cost volume. Filtering is restricted to these subvolumes, resulting in significant performance gains. We applied our method to disparity estimation from a stereo image pair and developed an occlusion handling method, which acts as a post-processing step that refines the disparity maps computed via filtering. Currently, we are exploring the use of slanted surfaces during the detection of salient subvolumes and we are implementing an efficient version of ACF for optical flow computation.
Source code is available at https://github.com/vclab/SVFilter-V1.
For technical details please look at the following publications