Depth Estimation Under Motion With Single Pair Rolling Shutter Stereo Images

Many methods have been proposed to process stereo matching for rolling shutter image pairs, they treat all pixels from an pair in identical way and require additional estimation approaches estimate motion states of the camera. However, a naturally diverse baseline lengths, are unstable one instantaneous input. In this letter, we present depth pipeline, which can robustly maps by alternating refining coarse levels fine levels. What is more, design novel cost volume building method adapts candidates change lengths pixels. We further demonstrate usability constructing new platform, outdoor evaluation dataset, comparing it with methods.