Human face structuring method and human face structuring system

Abstract

The invention proposes a human face structuring method and a human face structuring system wherein the human face structuring method comprises: obtaining a low-resolution human face image; according to the light distribution and detail distribution of the low-resolution human face image, adjusting the light distribution and detail distribution of each training set image in the training set; based on the each adjusted training set image, creating high-resolution image blocks through a neighborhood embedding method; and combining the high-resolution image blocks to obtain a high-resolution human face image. According to the technical schemes of the invention, a light-adaptation human face structuring method can be realized, which enables a low-resolution human face image to have the consistent light as each training set image so that their high frequency information is consistent and that high resolution reconstruction can be easily carried out, therefore, making the reconstruction of a high-resolution human face image more accurate.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular, to a face illusion method and a face illusion system. Background technique [0002] At present, face illusion refers to reconstructing the input low-resolution face image to obtain its corresponding high-resolution face image, thereby improving the performance of basic applications such as face recognition and face detection. In order to make better use of the relationship between corresponding low-resolution images and high-resolution image blocks, the researchers applied the super-resolution method of neighborhood embedding to face illusion and achieved good results. Based on the assumption that the low-resolution and high-resolution image blocks have the same local geometric features and neighborhood relationships, the corresponding high- and low-resolution dictionaries can be trained in advance, and the input low-resolution image can be used in the low-resolution dict...

AI Technical Summary

Problems solved by technology

[0003] However, this method of face illusion does not take into account the mismatch between the training set pictures and the input low-resolution images. For example, the low-resolution images have obvious highlights and shadows on the face due to strong light sources. distribution, while the lighting conditions of the training set pictures are relatively soft. Due to the inconsistency of high-frequency information under different lighting conditions, it is difficult to find highly similar image blocks in the training set pictures during the reconstruction process, resulting in the degradation of high-resolution reconstruction results.

Images