Method for correcting array multiple-view image of calibrated parallel cameras

Abstract

The invention discloses a method for correcting an array multiple-view image of calibrated parallel cameras. The method comprises the steps of: firstly, designing a uniform corrected camera physical coordinate system direction by using each camera external parameter in a pre-calibrated parallel camera to ensure that the rotary adjustable amount of a coordinate axis is minimum; then equally setting physical coordinate system original points of all corrected cameras at intervals on a public base line, establishing physical coordinate systems of all corrected cameras; and finally, uniformly setting internal parameters of the corrected cameras, calculating a correction conversion matrix of each camera and correcting multi-view images. According to the invention, vertical optical parallax can be effectively eliminated, and horizontal optical parallax equal distribution is realized, the rotating angle of the coordinate system in a correction process is small, the corrected image rotary distortion is small and accords with the vision habit of eyes, and thus the synthetic view image quality is remarkably improved, and the further processing of the subsequent processing link is facilitated.

Description

technical field [0001] The invention relates to a method for correcting multi-view images of parallel camera arrays, in particular to a method for correcting multi-view images of calibrated parallel camera arrays. Background technique [0002] The Free Viewpoint Television (FTV) system not only brings a strong sense of stereoscopic visual impact to the audience, but also provides the audience with the function of "moving and changing scenes" to interactively select the viewing position. In the FTV system, the multi-view video is captured synchronously by the camera array, sent to the server for processing, and finally the stereoscopic image is displayed on the autostereoscopic display. Every two adjacent viewing videos form a viewpoint video, and viewers can see different viewpoint videos at different positions, so that they can interactively choose the angle to enjoy the stereoscopic video. In multi-view video processing, multi-view image correction is a key issue. Since ...

AI Technical Summary

Problems solved by technology

Kang Y S, Lee C, Ho Y S. An efficient rectification algorithm for multi-view images in parallel camera array, 3DTV Conference: The True Vision-Capture, Transmission and Display of 3D Video, Istanbul, Turkey: IEEE, pp.61-64 , 2008 proposed a method of using camera calibration parameters to fit a common baseline and correct multi-view images of parallel cameras, but did not consider the image rotation distortion introduced in the image correction process

If the rotation distortion introduced by the multi-view image correction process is not strictly limited, on the one hand, it will cause visual discomfort, such as the upright object becomes tilted; The number of filled holes is large, which affects the correction efficiency

Images