Super-resolution image reconstruction method, super-resolution image reconstruction system, super-resolution image reconstruction device and storage medium

Abstract

The invention discloses a super-resolution image reconstruction method, a super-resolution image reconstruction system, a super-resolution image reconstruction device and a storage medium, and the method comprises the steps of inputting an original low-resolution image, preprocessing the image, and carrying out the bicubic interpolation on the input low-resolution image; and then, passing the preprocessed image through a sparse deep convolutional neural network containing pooling manifold constraints, mutually connecting and arranging the image data, learning the missing high-frequency detail information components in the image, and realizing the reconstruction of the high-resolution clear image. The method can effectively reduce the network parameter calculation amount and keep the important detail information of the image at the same time, and by inputting the original low-resolution image into the sparse deep convolutional neural network model trained by a set algorithm, the high-resolution image is realized by the super-resolution reconstruction.

Description

technical field [0001] The present invention relates to the technical field of image reconstruction, in particular to a face super-resolution image reconstruction method based on pooling manifold constraints and its system, device and storage medium. Background technique [0002] In recent years, my country has entered a period of rapid social development, and people pay more and more attention to the protection of personal privacy, property and personal safety. In order to maintain social order and citizen safety, video surveillance systems are widely used in traffic and security fields. For example, in the investigation and tracking of cases, the police need to obtain clear video surveillance quality in order to improve the detection rate. However, in actual monitoring, due to the resolution limitation of the imaging equipment of the video surveillance system, and the interference of factors such as the distance between the target imaged face and the camera, blur, noise, a...

AI Technical Summary

Problems solved by technology

However, once the deep neural network structure does not have a pooling layer, the image data is trained and tested in the form of a fully convolutional layer. Although the input and output feature maps between the various layers of the network remain unchanged, the amount of calculation of the neural network will increase. Large, then the time consumption will increase, resulting in negative impact

Although the existing research proposes to reduce the number of multiplication operations in the convolutional neural network by using the linear transformation of the neuron and the convolution kernel in the convolution operation or by modifying the network weight to perform neural network compression, once the neuron and convolution The linear transformation of the product kernel will make the network sparsity disappear, and the sparsity cannot be used to accelerate the network; the improved activation function is used in the full convolutional network to reduce the computational complexity between the neural network layers, but it does not reduce Occupancy and consumption of network memory and the amount of network parameters

Images