Image mosaicking method

Abstract

The present invention discloses an image mosaicking method. The method comprises the steps of: performing distortion correction of an image to be mosaicked according to a radial distortion coefficientof a lens; calculating an affine transformation matrix to perform image registration by identifying mark points; generating a fusion seam by employing a region growing method based on a pixel spatialdistance; calculating a mapping relation of a histogram consistency by employing a color histogram; performing brightness correction of a fused image; and finally, generating one contour layer and one detail layer corresponding to each layer by employing an image pyramid, and performing reconstruction to generate the fused image. The image mosaicking method employs the region growing method basedon the pixel spatial distance to generate the fusion seam of the image, and combines a traditional image super-resolution image pyramid reconstruction method to enhance an optimization effect; and moreover, the image mosaicking method is not limited by scenes and targets and can automatically calculate the fusion seam for targets with different scenes, depths of field and color spaces, so that the robustness and the generalization of image mosaicking are improved.

Description

technical field [0001] The present invention relates to the technical field of image stitching, in particular to an image stitching method. Background technique [0002] Image stitching is one of the earliest algorithms in the field of image processing. With the continuous attention of technologies such as video security and panoramic photography, the industry urgently needs new image stitching methods. Traditional image stitching algorithms can be divided into non-multi-scale decomposition image fusion algorithms and multi-scale decomposition image fusion algorithms. The former has a small amount of calculation, but the transition area is obvious after fusion, and the visual effect is poor; the latter transition area is soft, the visual effect is better, but the anti-noise ability is poor. [0003] According to different fusion methods, stitching algorithms can be divided into two categories based on optimal seam and smooth transition. The former calculates the best seam,...

AI Technical Summary

Problems solved by technology

The latter includes feathering-based fusion, gradient-based fusion, and pyramid fusion, etc., which are characterized by soft transitions, but have a large amount of calculation, poor anti-noise ability, and inaccurate combination of spatial features.

Images