Disparity Map Computation for Stereo Images using Compressive Sampling

Gade Narayna Harsha, Angshul Majumdar, and Rabab Ward


Stereo Vision, Disparity Map, Compressive Sensing


In this paper we propose a new technique for computing disparity maps. This paper is based on the popular correlation based disparity map computation approach. The task is to estimate a dense disparity map given two stereo images. Instead of computing the full disparity map, we only compute the disparity values for randomly selected epipolar lines. The output of this exercise is a sampled disparity map; from the sampled disparity map, the full map is reconstructed by exploiting the sparsity of the map in Fourier domain using Compressed Sensing techniques. Quantitative results show that our method of reconstructing disparity maps from sub-sampled data is only marginally infe- rior to the full disparity map computed by correlation based method.

Important Links:

Go Back