Image missing value recovery method based on non-rigid body locus base

Abstract

The invention discloses an image missing value recovery method based on a non-rigid body locus base. The image missing value recovery method utilizes a characteristic that the locus base can be customized to assume that the non-rigid body consists of r primitives and a rank of a formed image matrix is 3r+1, performs singular value decomposition on the image matrix to obtain a projection matrix, uses a property of the projection matrix to recover missing values of a row vector and a column vector, and uses the recovered missing values to replace the missing values of the image to perform multiple iterations until a position of the missing value is correctly recovered. In the process of recovery the missing values, the image missing value recovery method not only uses the property of the column vector, but also uses the property of the row vector, and equally treats all the images and characteristic points.

Description

technical field [0001] The invention belongs to the technical field of computer vision research, in particular to a trajectory-based missing value recovery method for non-rigid bodies. Background technique [0002] In computer vision, tracking the extracted feature points is one of the key steps in 3D reconstruction. However, due to light and occlusion, some of the feature points will be lost, which is one of the issues that must be considered in 3D reconstruction. In order to restore missing values, Jacobs in the paper "Linear fitting algorithm for motion structure recovery with missing points" (DJacobs.Linear fitting with missing data for structure-from-motion[J].ComputerVision and Image Understanding, 2008, 82(1) :57-81.), according to the characteristic that the rank of the image matrix is ​​4, the method of sub-matrix is ​​used to restore the missing value, but it is easily affected by the sub-matrix. One of the great advantages of the factorization method is that it ...

AI Technical Summary

Problems solved by technology

Based on the idea of ​​trajectory base, Gotardo et al., at the Computer Vision and Pattern Recognition Conference (P F U Gotardo, A M Martinez.Non-rigid structure from motion with complementary rank-3spaces.IEEE Conference on Computer Vision and Pattern Recognition[C],2011:3065 -3072.) Use the column vector fitting algorithm to reconstruct images with missing values, but only use the properties of column vectors

And it is based on the orthographic projection model, when the distance between the camera and the object is much greater than the depth of field of the object, the error is large

Images