Rapid and self-adaptive generation algorithm of intermediate viewpoint based on left and right viewpoint images

Abstract

The invention belongs to the image processing technique and particularly relates to a rapid and self-adaptive generation algorithm of intermediate viewpoint based on left and right viewpoint images. In order to accurately obtain the multi-viewpoint three-dimensional images of twin lens, the image at the real virtual viewpoint is obtained. The technical scheme comprises the following steps of: inputting left and right viewpoint images picked up by a camera to extract angular points of left and right images; extracting the angular point pair matched with the left and right views by using a method of normalization covariation; performing Delaunay triangulation on the left and right views according to the extracted accurate matching point by a point-to-point inserting method; acquiring an inverse transformation matrix to each triangle grid of the virtual viewpoint image; acquiring a point of each point in the triangle grid corresponding to the left and right viewpoint images, endowing the pixel value of the corresponding point to the point in the intermediate virtual viewpoint grid, and finally generating an image of the intermediate virtual viewpoint. The invention is mainly applied to image processing.

Description

technical field [0001] The invention belongs to the image processing technology and relates to the acquisition of multi-viewpoint stereoscopic images based on dual lenses. Specifically, the present invention relates to a fast and self-adaptive generation algorithm of the middle viewpoint based on left and right viewpoint images. Background technique [0002] The current virtual viewpoint generation scheme: [0003] 1. Eric Chen proposed a method for generating virtual viewpoint images. It first needs to know the position and direction parameters of the camera and the pixel depth information of the image, so as to establish the matching relationship between the images. Through linear interpolation, the current viewpoint can be generated. Virtual scene screen. Due to the need to provide camera parameters and depth information in advance, this method has great limitations in application. [0004] 2. McMillan proposed a method of generating new viewpoint images through reproj...

AI Technical Summary

Problems solved by technology

Due to the need to provide camera parameters and depth information in advance, this method has great limitations in application

[0004] 2. McMillan proposed a method of generating new viewpoint images through reprojection splicing. It assumes that the collected images are all shot with the same camera internal parameters, which is impossible, so there are limitations in the application

[0017] This algorithm is not aimed at the situation where two views are parallel. First, the images must be pre-transformed on the same plane, and then matched, and the matching method of dynamic programming is used, which has high complexity, and the images generated by interpolation have to be processed. After transformation to get the final virtual viewpoint image

Both pre-transformation and post-transformation will have deviations, resulting in an image that is not at the real virtual viewpoint

Images