Development of reference-based fast stereo matching system for the 3D endoscopic images

This paper presents the development of a fast depth extraction algorithm for 3D endoscopic data using a reference-based fast stereo matching method. Conventional stereo matching algorithms based on the integration of stereo and motion need a lot of time for computation. Firstly, the proposed algorithm selects reliable reference points based on the fusion of stereo and motion (successive frame), and then we perform unidirectional (L to R) stereo matching within one pair of stereo images with the reference points selected to save processing time. We used the system motion adaptively and the proposed system gives a fast as well as reliable matching result.