Computing the stereo matching cost with a convolutional neural network

We present a method for extracting depth informa-tion from a rectified image pair. We train a convo-lutional neural network to predict how well two im-age 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 er-ror rate of 2.61 % on the KITTI stereo dataset and is currently (August 2014) the top performing method on this dataset.