A Nonconventional Approach to the Conversion of 2D Video and Film Content to Stereoscopic 3D

This paper presents a nonconventional approach and method for converting naturalistic (non-computer-generated) two-dimensional (2D) video and film material to Stereoscopic 3D. Experimental evidence is presented to show the efficacy of color-based surrogate depth maps for automatic 2D-to-3D conversion aimed at small-screen applications. A semiautomatic 2D-to-3D conversion system (CRC-DMEG), also based on surrogate depth maps, for the conversion of video and film content for commercial projection on large cinema screens is then presented. A major advantage of CRC-DMEG is that it exploits the correlation between the 2D color images and the surrogate depth maps to allow for direct manipulation of the depth of objects in a scene. Another major advantage is that it allows for instant realization of depth details, such as raindrops, foliage, and textures found in carpets. This approach and method minimizes the labor-intensive work associated with the conventional method of rotoscoping. Moreover, the manual task of filling in disoccluded regions is also significantly reduced.