Stereo vision is the normal method to obtain the depth information from images. The problems encountered when applying well established algorithms to real time applications are due to the high computational load required. In this article, the authors address this issue by performing a region-based analysis which considers each pixel only once. Additionally, matching is carried out over statistical descriptors of the image regions. In this article, the authors present a new algorithm that has been specifically designed to solve some commonly observed problems which arise from other well known techniques. This algorithm was designed using a previous algorithm implemented by the authors. The complete analysis has been carried out over gray scale images. The results obtained from both real and synthetic images are presented in terms of matching quality and time consumption and are compared with other published results. Finally, a discussion of additional features related to the matching process is provided.
Pablo Revuelta Sanz, Belén Ruiz Mezcua, José M. Sánchez Pena, Jean-Phillippe Thiran, "Segment-Based Real Time Stereo Vision Matching Using Characteristic Vectors" in Journal of Imaging Science and Technology, 2011, pp 50201-1 - 50201-7, https://doi.org/10.2352/J.ImagingSci.Technol.2011.55.5.050201