Image Mosaicing Based on Binocular Cameras

This paper proposes a method for mosaicing images acquired simultaneously by binocular cameras fixed on different positions of a robot.It introduces adaptive non-maximal suppression techniques,implements Harris corner detection in the overlapped region of the image,and combines wavelet transformation with epipolar geometry to match image features efficiently.In image fusion,it also constructs a nonlinear transitional function,enabling the transition smoother in the overlapped region.Experimental results show the efficiency of the method is improved greatly compared with traditional approaches.