Image splicing method under non-concentric imaging condition

Abstract

The invention discloses an image splicing method under a non-concentric imaging condition. The image splicing method comprises the following steps: selecting a target candidate homography matrix conforming to a current scene by adopting a registration error-oriented cyclic selection mechanism; performing two-dimensional transformation coarse registration based on the target candidate homography matrix; densifying a sparse feature point registration error corresponding to the target candidate homography matrix by adopting an interpolation algorithm; obtaining a registration error of the whole overlapping region; carrying out error compensation, optimizing the image alignment, and increasing the image alignment area of the overlapping area, thus facilitating suture lookup; in addition, optimizing the suture line search by combining a feature point registration error constraint function, thus facilitating the stitching line searching function cost to be minimum, and finding the corresponding optimal stitching line, so as to realize perfect alignment of geometric structures of the images at two sides of the stitching line while the image content is not increased or deleted, and solve adouble image problem appearing in the complex scene image stitching process.

Description

technical field [0001] The invention relates to the technical field of digital image processing, and more specifically, relates to an image mosaic method under non-concentric imaging conditions. Background technique [0002] Image stitching technology is a difficult point in the field of computer vision. In the application of image stitching in complex scenes, the research on image stitching technology under non-concentric imaging conditions is the most important. Generally speaking, the image stitching process consists of two steps: image registration and image fusion. Image registration is the key to the whole image stitching process. The original image can be registered through two-dimensional space transformation, and the fusion operation can be performed directly after the image is aligned. This is the method adopted by most current stitching products, and it can correctly handle seamless image stitching under concentric imaging conditions. However, ideal stitching eff...

AI Technical Summary

Problems solved by technology

However, in this type of method, each loop iteration is a complete image stitching process, resulting in excessive time and memory consumption

Moreover, the seam search is directly based on the rigid transformation registration, and there will still be t

Images