Depth estimation is indispensable for a variety of tasks associated with the processing, editing, and displaying of stereoscopic 3D video. Stereoscopic video consists of two image sequences: one for the left eye and the other for the right eye. The pixel-by-pixel displacement between the left- and right-eye images is directly related to the depth of objects that the human visual system perceives.
Communications Research Centre Canada (CRC) has developed a fast algorithm for automatic depth estimation from stereoscopic video, which generates accurate and dense depth maps. The disparity map generator calculates disparity values per blocks, which can be as small as 2×2 pixels. It provides a dense disparity map that includes the maximum and minimum disparity values of the scene. It can also be used to provide vertical disparity if necessary.
The disparity map generator is implemented in C/C++ and is in the process of being implemented in VHDL. Its applications include text/graphic integration and editing in stereoscopic video, adjustment of displayed depth, and stereoscopic to multi-view video conversion.
original post: http://3dcinecast.blogspot.com/