Efficient image restoration method based on alternating direction method

Abstract

The invention discloses an efficient image restoration method based on an alternating direction method. The efficient image restoration method based on the alternating direction method includes: translating image restoration problems into low rank matrix complement problems; further dividing each low rank matrix into a plurality of partitioned matrixes by decomposing an original image into three low rank matrixes according to R / G / B channels; complementing each partitioned matrix; combining all the partitioned matrixes after being complemented so as to obtain corresponding complement matrixes; further obtaining a repaired image according an combination of the complement matrixes. The efficient image restoration method based on the alternating direction method facilitates parallel processing through partitioning, complements each partitioned matrix by using the alternating direction method, greatly improves image restoration speed, can not damage overall relevance of the original matrixes due to the fact that the plurality of the partitioned matrixes are obtained by dividing the three low rank matrixes according to relative independence of the three low rank matrixes, and facilitates improvement of restoration accuracy. The efficient image restoration method based on the alternating direction method not only can be used for image restoration, but also can be used in fields of object removal, image noise reduction and the like.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an efficient image restoration method based on an alternating direction method. Background technique [0002] Image inpainting is to restore the complete image through various methods based on the known parts of the image. At present, there are many methods, which can be divided into two categories: filling based on local information and approximation completion based on global information. [0003] Methods based on PDE algorithm (Partial differential equation, partial differential equation) and BP algorithm (Back Propagation, backpropagation) focus on local relations. The PDE algorithm tries to fill the lost area by deriving from the boundary of the lost area to the inside. The latter category first establishes a Markov random field (MRF) on the input image, and then performs global optimization through information exchange between adjacent nodes. Both types of algorithms assu...

AI Technical Summary

Problems solved by technology

Compared with the repair method based on local information filling, the calculation result of LRTC method is accurate, the repair precision is high, and the effect is good, but its convergence speed is very slow, resulting in low repair efficiency

[0005] To sum up, the existing image inpainting methods have a contradiction between inpainting quality and inpainting efficiency, and new methods are needed to achieve efficient and perfect image inpainting

Images