We present a method for extracting depth information from a rectified image pair. We train a convolutional neural network to predict how well two image patches match and use it to compute the stereo matching cost. The cost is refined by cross-based cost aggregation and semiglobal matching, followed by a left-right consistency check to eliminate errors in the occluded regions. Our stereo method achieves an error rate of 2.61 % on the KITTI stereo dataset and is currently (August 2014) the top performing method on this dataset.
@article{arxiv.1409.4326,
title = {Computing the Stereo Matching Cost with a Convolutional Neural Network},
author = {Jure Žbontar and Yann LeCun},
journal= {arXiv preprint arXiv:1409.4326},
year = {2015}
}
Comments
Conference on Computer Vision and Pattern Recognition (CVPR), June 2015