An Algorithm for Stitching Images with Different Contrast and Elimination of Ghost

Abstract

In image stitching, well performed stitching result is hard to achieve for images with large difference of contrast. This paper proposes a method based on histogram equalization by enhancing the contrast of stitching images. On the basis of image enhancement this method uses the algorithm of SURF (Speeded Up Robust Feature) feature points, K-Nearest Neighbor and bilateral matching method to match the feature points. Further, in order to get more stable and accurate homography, the method uses RANSAC (RANdom SAmple Consensus) algorithm. During the image fusion, as the traditional linear weighting method may generate ghost in the final image, this paper proposed a new method called four-section linear weighting method. Experimental results show that the method of the paper not only can realize image stitching with large difference of contrast, but also eliminate ghost phenomenon to a certain extent.