Panoramic Image Stitching Method Based on Bag-of-Words Model

Abstract

The invention relates to a panoramic image splicing method based on the bag-of-word model, which includes: preparing a trained ORB bag-of-word tree; performing ORB feature extraction for the image i in the image data set to be spliced, and assuming k to be the previous insertion into Splice the key frames of the structure diagram, conduct feature search on the two images through forward index, establish the correspondence between the features, and obtain the correspondence between the features between the two images; obtain the homography matrix between the two images; use random Sampling the maximum likelihood estimation algorithm to minimize the re-projection error and exclude outliers, and then obtain the inlier set of the corresponding image pair; calculate the overlap percentage of the image through the boundary matrix; calculate the overlap between the corresponding image pairs, when the image i When adding key frames to the splicing structure diagram, it is necessary to establish contact with the previous key frames through the homography matrix; detect loop closures; form loop closures; optimize the homography matrix; and fuse the image splicing structure diagrams.

Description

technical field [0001] The invention belongs to the fields of computer vision and digital image processing. Background technique [0002] Image mosaic technology is an important branch in the field of computer vision and digital image processing. It is a technology that seamlessly stitches more than two partially overlapping images to obtain higher resolution or wide viewing angle images. The two most critical links in image stitching are image registration and image fusion. For image fusion technology, the current method has little difference in time consumption and fusion effect, and it has become mature. But for image registration, because its registration time and effect directly affect the speed and success rate of image stitching, image registration has always been a hot spot in the research of image stitching. Common image registration methods are based on SIFT or SURF features to describe images. The advantages of these feature descriptors are rotation invariance, ...

AI Technical Summary

Problems solved by technology

However, most of the image stitching methods at this stage are based on frame-to-frame comparison methods. This method can receive good results when the number of images is small. Not applicable when higher

Images