A face image super-resolution secondary reconstruction method

Abstract

The invention discloses a face image super-resolution secondary reconstruction method, which comprises the following steps of firstly, carrying out face detection and target extraction on an acquiredmonitoring video to obtain a certain amount (20-30 frames) of target face image, carrying out quality evaluation on the obtained image based on an evaluation model, and preferentially selecting multiple frames (3-5 frames); secondly, carrying out super-resolution reconstruction on the result to synthesize a plurality of frames of images into a high-quality virtual image; constructing a face imagesuper-resolution reconstruction model MRES based on the convolutional neural network CNN again, wherein the model is used for learning the mapping relation between the high-resolution sample image andthe corresponding low-resolution image and is based on an inception structure for removing a pooling layer, a residual error learning idea for reducing learning difficulty is adopted, a multi-scale aggregation module capable of comprehensively extracting characteristics is used, and a deconvolution layer is added to replace an interpolation operation; and finally, training the second step by using the training model in the third step to obtain a high-resolution face image. According to the method, the reconstruction effect can be improved within the controllable training time.

Description

Technical field [0001] The invention relates to image processing and computer vision technology, in particular to a method for super-resolution secondary reconstruction of a face image. Background technique [0002] In recent years, the safety of public places has always been a concern, especially in densely populated places. The state proposed the Skynet Project for urban management and public security prevention and control. Therefore, future monitoring systems will become more and more popular. [0003] Due to problems such as the shooting distance and angle and the resolution of the monitoring equipment itself, the generated face images are sometimes blurred, incomplete, noisy, and small in size, which affects recognition. [0004] Super-resolution reconstruction is to process one or more low-resolution images through software methods to obtain high-resolution images to facilitate recognition. [0005] The commonly used super-resolution reconstruction techniques can be summarize...

AI Technical Summary

Problems solved by technology

[0003] Due to problems such as shooting distance and angle and the resolution of the monitoring equipment itself, the generated face images are sometimes blurred, incomplete, noisy, and small in size, which affects recognition

Images