Stereo Vision is one of the fundamental problems of computer vision and one of the oldest and heavily investigated areas of 3D Vision. Recent... Show moreStereo Vision is one of the fundamental problems of computer vision and one of the oldest and heavily investigated areas of 3D Vision. Recent advances in stereo matching methodologies have allowed researchers to develop several stereo vision techniques. But it has been observed that most of these methodologies mainly focus on only one aspect, either the execution time or the quality of the generated depth map. While attempts to maximize the quality of the depth map have been successfully achieved, it is however observed that the execution time has increased signi cantly, thereby rendering them un t for real time implementation. In this thesis, an attempt to solve the above said problem is presented, in which we combine the e ects of two important stereo matching concepts, one attempting to increase the depth map quality, while the other being that of having an algorithmic structure so that in future, it can be used for real-time implementation. By applying hardware-friendly methods to the framework of segment based stereo matching, the method is said to achieve a good quality depth map. This work also discusses the various challenges which it takes care of plus the study of performance compared to existing methods. M.S. in Electronical Engineering, July 2012 Show less
