Integrating LIDAR into Stereo for Fast and Improved Disparity Computation

TLDR: This paper proposes to integrate LIDAR data directly into the stereo algorithm to reduce false positives while increasing the density of the resulting disparity image on textureless regions, and demonstrates with extensive experimental results that the disparity estimation is substantially improved while speeding up the stereo computation.

Abstract: The fusion of stereo and laser range finders (LIDARs) has been proposed as a method to compensate for each individual sensor's deficiencies - stereo output is dense, but noisy for large distances, while LIDAR is more accurate, but sparse. However, stereo usually performs poorly on textureless areas and on scenes containing repetitive structures, and the subsequent fusion with LIDAR leads to a degraded estimation of the 3D structure. In this paper, we propose to integrate LIDAR data directly into the stereo algorithm to reduce false positives while increasing the density of the resulting disparity image on textureless regions. We demonstrate with extensive experimental results with real data that the disparity estimation is substantially improved while speeding up the stereo computation by as much as a factor of five.