PROJECT TITLE :

Depth Reconstruction From Sparse Samples: Representation, Algorithm, and Sampling - 2015

ABSTRACT:

The fast development of 3D technology and computer vision applications has motivated a thrust of methodologies for depth acquisition and estimation. However, existing hardware and software acquisition methods have restricted performance due to poor depth precision, low resolution, and high computational value. In this paper, we have a tendency to present a computationally economical technique to estimate dense depth maps from sparse measurements. There are 3 main contributions. 1st, we tend to offer empirical evidence that depth maps will be encoded much a lot of sparsely than natural pictures using common dictionaries, like wavelets and contourlets. We tend to also show that a combined wavelet-contourlet dictionary achieves higher performance than using either dictionary alone. Second, we propose an alternating direction methodology of multipliers (ADMM) for depth map reconstruction. A multiscale warm begin procedure is proposed to hurry up the convergence. Third, we tend to propose a 2-stage randomized sampling theme to optimally select the sampling locations, so maximizing the reconstruction performance for a given sampling budget. Experimental results show that the proposed method produces high-quality dense depth estimates, and is sturdy to noisy measurements. Applications to real data in stereo matching are demonstrated.