基于多平面的图片拼接

Image stitching using multi warping matrices

图片拼接是一种可以将多幅图片整合成单张全景图的技术,由于图片间视差导致的拼接错误通常是图片拼接中的难点部分。前人使用的很多拼接方法都是基于单个变换矩阵得到的,这种方法在非平面的拼接中往往无法得到正确的结果。基于上述问题,本文章采用了一种多平面拼接的方法,通过将图片内容分割成不同的平面,对每一个平面采用统一的变换矩阵,并采用图片缝切技术降低拼接造成的噪声,从而得到最后的拼接结果。在实验部分,由于采用了更加接近真实环境的多平面拼接假设,所以得到的结果具有更少的畸变效果。

This paper tries to construct image panoramas using multi warping matrices. A critical problem in image stitching is the parallax caused by the motion of the camera. There are many previous works inthis field , and a large number of them only used a single projective warp , which may not be correct when the contents are not planar. This paper uses a method which divides the image into different parts inwhich contents are nearly planar. Then by combining all parts together and using a seam cut can produce the final result. Its assumption is more close to the real model , and the result shows thatgood quality images are created.