Fast and low complexity image stitching method on mobile phones

Abstract

Dealing with image stitching on mobile phones has always been a major image processing problem in the research of computer vision. Image stitching methods employed on PC are well studied, and many successful applications have been received on the products of panorama. However, image stitching has more challenge on mobile phones, which have limited processing power and memory. In this paper, a fast and low complexity image stitching method on mobile phones was proposed, and the stitching pipeline is formulated as a sequential processing problem. The coarse matching with SIFT features was first got in our proposed method, followed with the precise matching result that got by RANSAC, and a projective transformation matrix can be obtained. The mask fusion method is then employed to compose input images into panorama. In this way, the proposed method can get panorama with less computational resources and preserve more reality image structure on mobile phones. Experiments on realistic images validate the effectiveness of this method on mobile phones in terms of speed and visual effects when comparing with the autostitch method.