Face-based defected image inpainting method and system

Abstract

The invention provides a face-based defected image inpainting method and system. The method comprises the following steps of acquiring pixels of a whole face from a defected face image; retrieving a plurality of face images, similar to the defected face image, from a face image database through sparse representation according to the pixels of the defected face image; extracting the best-matched defected portion from the face images according to a defect area; and intercepting the best-matched defected area from the best-matched face image, and fusing the defected portion by using a Laplace equation so as to obtain a repaired image. According to the method provided by the embodiments of the invention, the complexity of data calculation during image retrieval is reduced by using sparse representation, so that the efficiency is increased; and meanwhile, the image of the defect area is subjected to inpainting by using spectrogram information of the image.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a face-based image defect inpainting method and system. Background technique [0002] Face recognition is the most natural, non-destructive, user-friendly method of biometrics. Most existing systems can only successfully recognize faces under certain constraints. In some special cases, the recognition results will be affected by images, such as lighting, facial expressions, postures, viewing angles, especially sunglasses, beards or occluders in masking phenomena. [0003] The face recognition uses the entire complete face image as the recognition object. When the recognized face image information is incomplete, the general recognition method will not be applicable. [0004] Currently commonly used inpainting methods for damaged images include inpainting algorithms based on Poisson equations and inpainting algorithms based on Markov fields. However, traditional inpaint...

AI Technical Summary

Problems solved by technology

[0006] (1) Unable to deal with the lighting of the area to be matched and the target area, and the obvious change of the color difference, etc.

[0007] (2) When applied to the human face, due to the lack of original face image information, under the framework of the traditional inpainting algorithm, it is not very good to find the small area to be matched for the target inpainting

Images