Depth manipulation using disparity histogram analysis for stereoscopic 3D

Abstract

The importance of post-production for stereoscopic 3D is increasing rapidly. In particular, depth manipulation is essential, as there are many situations in which the captured depth requires further adjustment. Nonlinear disparity mapping has been a popular choice for efficient depth manipulation. However, most existing work requires users to have a deep understanding of how stereo works. This paper proposes a novel and very intuitive-to-use nonlinear disparity mapping technique. A commonly used multirigging technique inspired this work. Specifically, our method creates multiple depth layers using the Gaussian Mixture Model (GMM) and a histogram analysis. The depth position and volume are then manipulated with simple parameters at each layer individually, achieving complex nonlinearity in terms of depth control. The employed optimization scheme ensures the preservation of the original depth order. A user study shows that our method is very easy to use and simple to control. We demonstrate the versatility of our method with various practical applications.