Image stitching and ghost elimination based on Shape-Preserving Half-Projective warps

This paper presents a fast image stitching method based on the Shape-Preserving Half-Projective warps. The method includes three principal steps: Firstly, the feature points are extracted and matched using the SIFT algorithm; Secondly, image registration is completed by estimating the converting relationship between images with the RANSAC and the Shape-Preserving Half-Projective Warps. Finally, seamless image fusion is obtained by using the dynamic programming of optimal stitching line search and the multi-resolution spline algorithm, which can reduce the effects of seams and illumination variations and eliminate ghost caused by moving objects. Experiments demonstrate that this method has superior quality of stitching.