Image homography matrix calculation method, image homography matrix calculation device, mobile terminal and storage medium

Abstract

The invention provides an image homography matrix calculation method, an image homography matrix calculation device, a mobile terminal and a storage medium. The method comprises the steps of: obtaining image information of a first image and a second image, and obtaining at least one pre-set homography matrix stored locally; respectively performing image transformation on the first image accordingto each pre-set homography matrix, so that corresponding transformed images are obtained, and combining all the transformed images with the second image, so that multi-channel information is obtained;and, inputting the multi-channel information into a convolutional network, so that multiple transformed homography matrixes are output, and calculating the multiple transformed homography matrixes according to a pre-set calculation rule, so that a target homography matrix is obtained. According to the image homography matrix calculation method, the image homography matrix calculation device, themobile terminal and the storage medium provided by the invention, design of the multiple transformed images is obtained through image transformation on the first image; therefore, graphic informationinput into the convolutional network is abundant; the homography matrix calculation precision is improved; furthermore, the overlapping degree of the first image and the second image can be improved through transformation of the multiple pre-set homography matrixes; and, influence of relatively large visual angle difference of the two images can be reduced.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image homography matrix calculation method, device, mobile terminal and storage medium. Background technique [0002] With the development of the times and the continuous improvement of people's living standards, the use of cameras is becoming more and more frequent. When the camera captures images of the same scene from different perspectives, there is a unique optimal homography matrix between two images. This matrix can transform one of the images to coincide with another image, and realize complex computer vision tasks such as image stitching, fusion, dual-camera calibration, and depth estimation. Therefore, how to calculate the homography matrix of two images of the same scene , has always been the focus of image processing technology. [0003] The method adopted in the existing image homography matrix calculation method is to directly input two images into the ...

AI Technical Summary

Problems solved by technology

[0004] In the existing image homography matrix calculation method, because two images are directly input to the convolutional network, the amount of image overlapping information in the convolutional network is too small, resulting in low accuracy of the calculated homography matrix, and Due to the large deviation of the viewing angles of the two images, it is difficult to cover the overlapping part of the two images with a small convolution kernel, making it difficult for the convolutional network to capture the correspondence between the images, which further makes the calculated homography Matrix is ​​less accurate

Images