An Unsupervised Stereo Image Redirection Method

Abstract

The invention discloses an unsupervised stereo image redirection method. The method includes: using a multi-level attention generation module to obtain an attention map of a stereo image; using the correlation between viewpoints between left and right images to construct a viewpoint synthesis loss; Use the consistency of stereo images before and after redirection to construct stereo loop consistency loss; build a total loss function based on viewpoint synthesis loss and stereo loop consistency loss, and train unsupervised stereo image redirection through the total loss function to obtain redirection post-stereoscopic image. The invention utilizes the unsupervised deep learning method, adopts the multi-level attention generation module to extract high-level features, extracts the information of the salient area, and utilizes the unsupervised viewpoint synthesis loss and the stereo cycle consistency loss to ensure the geometric structure and depth information of the stereo image, Implement stereo image redirection.

Description

technical field [0001] The invention relates to the technical fields of image processing and stereo vision, and in particular, to an unsupervised stereo image redirection method. Background technique [0002] Stereoscopic images can provide an immersive visual experience and have received extensive attention from industry and academia. With the growth of different stereoscopic display devices, there is a need to display stereoscopic images and videos on devices of different resolutions and target aspect ratios. Stereoscopic image redirection technology aims to intelligently process the multimedia content of the display device to adapt it to different sizes of screens, and can be widely used in virtual reality, human-computer interaction and other fields. [0003] Currently, 2D image retargeting methods are divided into discrete methods and continuous methods. Discrete methods change the original size by removing or interpolating pixels that contribute the least energy to t...

AI Technical Summary

Problems solved by technology

[0006] The methods in the prior art use 2D image redirection technology to process stereoscopic images, ignoring the disparity information of stereoscopic images, which will lead to inconsistency in the deformation of significant areas, and further weaken the depth perception of 3D scenes; the existing discrete-based stereoscopic image redirection methods may result in distortion of the shape and content of stereoscopic images, while continuous-based stereoscopic image retargeting methods often result in parallax distortion of stereoscopic images

Images