An image adaptive fusion method

Abstract

The invention relates to an image adaptive fusion method, belonging to the image processing field. The invention converts image fusion in multi-imaging system into pixel interpolation method in new coordinate system, The left pixel region of the new image is completely taken from the pixel information in img1, the right region of the new image is completely taken from the img2 image data after thecoordinate information is transformed, and then the pixel information of the overlapping region in the middle is evaluated and optimized by an adaptive method, so that the pixel information of the overlapping region is fuller. The invention introduces the idea of cosine distance discriminant analysis into image fusion technology, Compared with traditional technology, it greatly reduces the possibility of image distortion after fusion, The robustness is better, which can effectively correct the possible non-uniformity of metrics when each pixel is selected, automatically select and plan the appropriate imaging point set of each camera, the result of image fusion is smoother, at the same time, the image detail information is fuller, and the overall effect is more realistic.

Description

technical field [0001] The invention belongs to the field of image processing, and relates to an image adaptive fusion method based on cosine distance discriminant analysis. Background technique [0002] At present, machine vision has become a hot topic in the industrial field. The introduction of vision in the industrial production site can better complete the identification, detection, packaging and handling of workpieces. However, when obtaining the visual information of the workpiece, some production items need to be identified. The field of view of the workpiece is relatively large, and a single camera cannot well meet the actual needs, so a multi-camera imaging system with a parallel optical axis structure (such as the attached figure 1 ), and finally use the image stitching algorithm to get the required image data. [0003] The core of the image stitching algorithm is mainly image registration and image fusion. The method of image registration has been relatively mat...

AI Technical Summary

Problems solved by technology

[0004] The existing image fusion methods are mainly divided into two types: a simple method is to first attach the figure 2 The right image img2 is transformed into a new image by the transformation matrix H, and then the left image img1 is directly added to the new image, so that there will be obvious stitching seams after splicing; another method is to firstly convert the right image img2 Transformed into a new image by the transformation matrix H, the target image is divided into three parts during the fusion process, the leftmost part is completely taken from the data in img1, the right part of the overlapping area is completely taken from the transformed image of img2, and the middle part The weighted average of the two images is taken for the overlapping part, which may lead to the loss of details in the overlapping area, and there will also be certain stitching seams

Images