Sparse reconstruction for disparity maps using combined wavelet and contourlet transforms

Disparity estimation is a key component in 3D image processing, yet dense estimation is a computationally intensive task. In this paper, we propose to estimate the dense disparities from a small set of spatial measurements. Observing that disparity maps mainly contain contours and smooth regions, we formulate the problem as a sparse reconstruction problem using a combined wavelet and contourlet bases. We show that the combined transform yields better reconstruction results than existing methods.