A video mosaic method based on optical flow

Abstract

The invention discloses a video mosaic method based on optical flow, which comprises the following steps: step S1: reading and preprocessing left and right frame images; 2, calculating the corresponding common area of the left and right frame image, and estimating the dense optical flow of the common area; Step S3: clustering the dense optical flow of the common area into three categories; Step S4, estimating the extension of the optical flow of the non-common area, and projecting the non-reference frame onto the reference frame to form a preliminary panorama according to the optical flow; 5,detecting and correcting that conceal area in the preliminary panorama; Step S6: returning to step S1, and reading the next pair of left and right frame images. The method of the invention aims at thesynchronous monitoring video with fixed camera direction, and can accurately find the corresponding relationship between the common areas of the left and right frame images through the steps, therebyobtaining the panoramic video with good effect, and greatly weakening the blurring and twisting effect on the moving object.

Description

technical field [0001] The present invention relates to the technical field of video stitching, in particular to an optical flow-based video stitching method aimed at optimizing blur and distortion effects of moving objects in public areas. Background technique [0002] With the development of technology, people have higher and higher requirements for video, and one of the most important aspects is the field of view. The field of view of video captured by a single camera is too narrow, and the content that can be presented is relatively small, which cannot meet people's needs. Therefore, people can only adjust the focal length to obtain a wider field of view, but the adjustment of the focal length will cause the picture to become blurred and the resolution to decrease. Later, with the development of the microelectronics industry, people invented wide-angle lenses that can obtain a larger field of view. However, wide-angle lenses are generally expensive and difficult to popu...

AI Technical Summary

Problems solved by technology

With the remarkable results of image stitching, people gradually apply this method to video. For fixed multiple cameras, since the transformation matrix H between videos remains unchanged, it is only necessary to calculate the transformation matrix H at the beginning. This transformation matrix H can be used all the time in the subsequent video frames. This method saves time to a great extent and the splicing speed is faster, but it does not take into account the moving objects appearing in the video, which will affect the optimal image between two frames. Transformation matrix, so it often causes serious blurring effects around moving objects; another video stitching technology will consider calculating the transformation matrix for each frame, so that moving objects can be taken into account, but sometimes the transformation matrix of two adjacent frames changes too much The conference causes jumping between frames and affects the visual effect

[0006] In 2015, Jiang W et al. proposed a spatial-temporal local mapping method to optimize the effect of moving objects. At the same time, the fusion part considered the information of the front and rear frames and proposed 3D suture fusion. This method can effectively optimize moving objects. Stitching effect, but the effect is poor for situations where the difference in illumination between the left and right frames is too large, and there are few feature points that can be extracted from moving objects.

Images