An Enhancement Method of Obtaining Interest Points in Binocular Vision

摘要

The key element of obtaining the depth value by stereo matching in binocular vision is interest points. The more interest points, the more computational burdens normalized cross-correlation (NCC) stereo correspondence has. The paper proposes a method based on image enhancement realized among indoor or urban canyon which aims at making a better trade-off between the quantity of interest points and the computational burdens. In this paper, for some cases in normalized cross-correlation (NCC) stereo correspondence period, two times more interest points were obtained by using the proposed image enhancement method than function "stretchlim" and "imadjust". The proposed method will be carried out by various platforms with binocular vision such as Unmanned Aerial Vehicles (UAVs) and other ground robots to realize obstacles detection and avoidance.