Stereo matching based on colour invariants in RGB orthogonal colour space

摘要

Stereo matching is a key task in computer vision, such as 3D reconstruction and image registration. This paper presents a new stereo matching method based on colour invariants C_λ and H according tothe Kubelka-munk theory in RGB orthogonal colour space, using SURF feature vectors and Euclidean distance to match colour images. The experiments were carried out by using the ALOI database and images captured in Beijing Forestry University campus and the results show that the proposed method has a strong performance about evaluation criteria compared with SIFT, SURF algorithm in grey level and CSURF algorithm approximated to traditional CSIFT with C_λ and Hcharacters. Moreover, we illustrate indirectly that the RGB orthogonal colour space has colour invariance and why almost all the improved methods in the application of point extraction do not use C_λ character, only H property.