Bundling adjustment method for image splicing, storage medium and computing equipment

Abstract

The invention relates to a bundling adjustment method for image splicing, a storage medium and computing equipment, and the method comprises the steps: S1, carrying out the initialization setting: presetting a transformation matrix between corresponding point pairs on an image to be spliced, enabling the initialization cumulative gradient to be zero, and presetting a threshold value of an error; S2, constructing an error function to calculate an error function value of the transformation matrix, and constructing a gradient function of the error function value of the transformation matrix to the virtual target transformation matrix; S3, calculating a virtual target transformation matrix, and calculating the gradient of the error function value of the transformation matrix on the virtual target transformation matrix according to the gradient function; S4, updating the cumulative gradient and updating the transformation matrix according to the gradient of the virtual target transformationmatrix; S5, calculating an error function value of the updated transformation matrix; and S6, when the error function value of the updated transformation matrix is smaller than a threshold value, outputting the updated transformation matrix, otherwise, updating the number of iterations, and repeating the steps S3-S6 by using the updated transformation matrix.

Description

technical field [0001] The invention relates to the technical field of computer vision, and in particular to a bundle adjustment method, a storage medium and a computing device for image splicing. Background technique [0002] Image stitching belongs to the field of computer vision. When using a camera to shoot the actual scene, it is often impossible to achieve a wide field of view and a high resolution at the same time, so image stitching technology is produced, that is, using multiple cameras to shoot different areas of the actual scene, or using the same The camera takes multiple shots of different areas of the actual scene, and then uses the computer for post-production synthesis, so as to meet the requirements of wide field of view and high resolution at the same time. [0003] In post-compositing, the key to stitching is to find the transformation relationship between two adjacent images in the actual scene, and the clues lie in some corresponding point pairs in the ...

AI Technical Summary

Problems solved by technology

[0011] In order to solve the technical problems of complex calculation, low iteration speed and local oscillation at the optimum point in the existing bundle adjustment method, the present invention provides a bundle adjustment method and storage medium for image mosaic based on the Nesterov accelerated gradient method And computing equipment, when iteratively obtaining the optimal solution of the transformation matrix H, not only the impact of the previous iterations on this iteration, but also the impact of the next iteration will be considered, thereby ensuring the reliability of the iteration, That is, it will not stay at the local optimal solution, but can converge to the global optimal solution, and also ensure the effectiveness of iteration, that is, it will not oscillate near the optimal solution, but efficiently iterate to the optimal solution

Images