Real-time Hybrid Stereo Vision System for HD resolution disparity map(BMVC2014)

Stereo vision systems have been researched for decades and constitute the traditional method for extracting three-dimensional (3D) real-time depth information from images using sensors. However, passive stereo vision systems show a significant error in processing untextured regions, which are frequent in indoor environments. Also, modifications for processing in real time make error by using approximate algorithms. In this paper, we propose a hybrid stereo matching system that combines active and passive stereo vision. In order to implement it in a field-programmable gate array (FPGA) for real-time processing, we modify the existing stereo matching algorithm. Our system shows a significant improvement over current systems in processing untextured regions, and accurately calculates depth for a 1280 x 720-resolution image at 60fps in indoor environments. Paper: http://www.bmva.org/bmvc/2014/papers/paper074/index.html