Seamless registration of dual camera images using optimal mask-based image fusion

摘要

In an image fusion, misalignment errors are generated in the boundaries between objects and background. To solve this problem, this paper presents an image fusion method using the optimal mask that contains main objects. The proposed method consists of three steps: i) pre-processing and global warping to register a tele-view image to a wide-view image, ii) generating seam-based optimal mask to detect the noticeable region, and iii) image fusion using the mask and local warping based on motion estimation. This method provides high quality fusion results without misalignment. For this reason, the proposed method can apply in various areas that need the image synthesis.