Improved human face super-resolution reconstruction method based on nearest feature line manifold learning

Abstract

The invention discloses an improved human face super-resolution reconstruction method based on nearest feature line manifold learning, and the method comprises the steps: carrying out the discrimination of the condition of an extrapolation line of a connection line with a projection point being located between two sample points based on a conventional human face super-resolution reconstruction method based on nearest feature line manifold learning, i.e., searching the sample point which is nearest to the projection point from the two sample points to replace the projection point when the sum of the Euclidean distances between the projection point and the two sample points is W times greater than the Euclidean distance between the two sample points; forming a to-be-screened point set. Therefore, the restriction on the projection point enables the correlation between the projection point and the sample points to be stronger, improving the expression capability of newly obtained sample data for an input low-resolution image block to a great extent, avoiding the introduction of detail information that an original image does not comprise as much as possible, and improving the reconstruction effect of a low-resolution image.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an improved human face super-resolution reconstruction method based on nearest feature line manifold learning. Background technique [0002] The 2016 government work report emphasized: "Innovate the comprehensive management mechanism of social security, promote the construction of social security prevention and control system with the support of informatization, punish illegal and criminal acts in accordance with the law, severely crack down on violent and terrorist activities, and enhance the sense of security of the people"; currently Among the many security measures, video surveillance and image processing technology are playing an increasingly important role in preventing and combating crimes. However, according to statistics, the quality ratio of surveillance images obtained during the day is as high as 60%, and at night it is as high as 95%. %, therefore, how to re...

AI Technical Summary

Problems solved by technology

[0002] The 2016 government work report emphasized: "Innovate the comprehensive management mechanism of social security, promote the construction of social security prevention and control system with the support of informatization, punish illegal and criminal acts in accordance with the law, severely crack down on violent and terrorist activities, and enhance the sense of security of the people"; currently Among the many security measures, video surveillance and image processing technology are playing an increasingly important role in preventing and combating crimes. However, according to statistics, the quality ratio of surveillance images obtained during the day is as high as 60%, and at night it is as high as 95%. %, therefore, how to reconstruct a high-quality recognizable face image based on the original low-quality suspect’s face image has become an urgent need for video surveillance

[0006] Although the above method greatly expands the expression ability of sample data, it lacks the necessary constraint information when selecting the nearest neighbor projection point, and introduces detailed information that does not exist in the original image, resulting in unsatisfactory image reconstruction effect.

Images