Computer vision problems are much larger than ever before. One way to satisfy demand is to design more efficient and clever algorithms that optimize computations in an existing processor. Such fast algorithms are becoming increasingly important for tele-reality, interactive media and visual serving applications. This book presents fast and reliable algorithms for dense stereo matching – for every point on the image – and optical-flow estimations using a general language, such as C, rather than dedicated hardware implementation.
Techniques described are: fast algorithms for similarity measure, use of subregioning technique to expedite similarity calculation, multiresolution scheme, fast 3-D surface technique for stereo matching, and stereo matching using warping. Fast Algorithms for Stereo Matching is a useful reference for academics, professionals, researchers, practitioners, and advanced graduate students in the areas of computer vision, digital photogrammetry, and 3-D video coding.