Depth sparse representation algorithm-based high-resolution-rate human face image synthesis method

Abstract

The invention provides a depth sparse representation algorithm-based high-resolution-rate human face image synthesis method. The main content comprises: a dictionary learning algorithm, a depth sparserepresentation algorithm, training and testing; and the process is as follows: taking low-resolution-rate test data as input; learning a k-grade depth dictionary in source and target domains for a low-resolution-rage human face image and a high-resolution-rate human face image; learning all the k-grade depth dictionaries by the corresponding sparse representation; learning the first-grade source(low resolution rate) and a target (high resolution rate) domain dictionary; learning the second-grade low resolution rate and high resolution-rate image dictionaries; and learning conversion among the final representation so as to acquire synthetic output of the given image. By the method, the resolution rate of the given low-resolution-rate input is enhanced, the image quality is improved and agood identification result is generated; furthermore, the time used by the algorithm is extremely short and the effectiveness and the usability of the algorithm in the low-resolution-rate human face identification application are shown.

Description

technical field [0001] The invention relates to the field of image synthesis, in particular to a high-resolution human face image synthesis method based on a deep sparse representation algorithm. Background technique [0002] With the rapid development of information technology, electronic image monitoring has been widely used in many fields, such as traffic monitoring, military inspection, security precautions in public places, etc. The clarity and quality of electronic monitoring images will directly affect the actual application effect of the monitoring system. Therefore, only by clearing these blurred images and generating high-resolution images through technical means can the images provide effective information. For example, in the field of security protection, when security personnel or criminal investigators use surveillance video and other shooting pictures to lock the target person, they can use the enlarged low-resolution face image as input and use the high-resol...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the processing time is too long and the restored image is not good, the purpose of the present invention is to provide a high-resolution face image synthesis method based on a deep sparse representation algorithm, which uses low-resolution test data as input, and for low-resolution resolution face images and high-resolution face images, learn k-level depth dictionaries in both the source and target domains, and the corresponding sparse representations also learn all k-level depth dictionaries, and then learn the first level source (low-resolution ) and the target (high-resolution) domain dictionary, then learn a second level of low-resolution and high-resolution image dictionaries, and finally learn the transformation between the final representations to obtain a synthetic output for a given image

Images