Method for estimating optical flow directed at large displacement

Abstract

The invention discloses a method for estimating optical flow, directed at large-scale movement, aiming at computing dense pixel corresponding relationship among images containing large displacement and large-scale varying scenarios. The method includes the following steps: firstly, pre-processing input images; then constructing an optical flow control point set, and computing a control optical flow vector and a control point state matrix, and constructing an initial optical flow vector and an optical flow offset vector; and finally, realizing iterative optimization of the optical flow under the framework of discrete optimization. According to the invention, the method can effectively process large displacement and large-scale varying scenarios in the estimation of optical flow, noticeably increases the precision of optical flow estimation of large-scale movement, and has noticeable increase in computing efficiency compared with prior art.

Description

technical field [0001] The invention belongs to the technical field of image information processing, in particular to an optical flow estimation method for large-scale motion. Background technique [0002] Optical flow estimation refers to the process of calculating the dense pixel correspondence between two or more images. Optical flow estimation is a basic problem in the field of image information processing research. Optical flow estimation has important applications in many computer vision tasks such as motion estimation, video compression, image registration, view interpolation, 3D reconstruction, etc. In the past one or two decades, optical flow estimation methods have been widely studied, and the performance of optical flow estimation has also made great progress. [0003] At present, the more popular optical flow estimation method is the classic deformation optical flow estimation method (T.Brox, A.Bruhn, N.Papenberg, and J.Weickert, "High accuracy optical flow esti...

AI Technical Summary

Problems solved by technology

This method can improve the performance of large-displacement optical flow estimation to a certain extent, but its accuracy is limited to the pixel level and cannot achieve sub-pixel optical flow estimation.

Motion detail preserving optical flow estimation method (L.Xu, J.Jia, and Y.Matsushita, "Motion detail preserving optical flow estimation," IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(9):1744-1757, 2012) can handle Optical flow estimation for complex large-scale motion, but the operating efficiency of this method is much lower than that of classical deformable optical flow estimation methods

Images