Deep learning super-resolution reconstruction method based on residual sub-images

Abstract

The invention discloses a deep learning super-resolution reconstruction method based on residual sub-images; residual sub-images are effectively combined with deep learning method based on convolutional neural network, super-resolution reconstructed images are clearer, and reconstruction speed is higher. By increasing the depth of convolutional neural network, a network model acquired by learning has higher nonlinear expression capacity and image reconstructing capacity; in addition, by introducing residual sub-image process, preprocessing based on traditional interpolation algorithm is removed, and fuzzy effect due to the interpolation algorithm is avoided. By making ingenious use of residual sub-images, it is possible to transfer deep learning convolutional operation from high-resolution space to low-resolution space, and accordingly reconstruction efficiency of super-resolution algorithm is increased at the premise of improving super-resolution reconstruction effect.

Description

technical field [0001] The present invention relates to the field of computer image processing and artificial intelligence technology, in particular to a deep learning super-resolution reconstruction method based on residual sub-images. Background technique [0002] Image super-resolution is the process of reconstructing high-quality high-resolution images from low-resolution images. It has broad application prospects in the fields of video compression and transmission, medical image-aided diagnosis, security monitoring, and satellite imaging. There are two evaluation criteria for image super-resolution: (1) image reconstruction effect, the goal is to restore the high-frequency information of the image, improve the quality of the image, and improve the visual effect of the reconstructed image as much as possible; (2) image reconstruction Efficiency, the goal is to increase the reconstruction speed as much as possible while ensuring the reconstruction effect. The current mai...

AI Technical Summary

Problems solved by technology

However, the image super-resolution reconstruction effect of this patent is average, and the reconstruction efficiency can be further improved. Its specific shortcomings are: (1) This patent uses bicubic interpolation as preprocessing, which is actually a combination of traditional methods and machine learning methods. In this combination, the reconstruction effect will be affected by traditional interpolation methods, such as blurring effects

(2) This patent only uses a 3-layer network structure, and its nonlinear feature representation ability and image reconstruction ability are limited

(3) The training data of this patent are low-resolution images and high-resolution images. The low-frequency information of the high-resolution images will also be reconstructed during the training process of the network, so there is no specific reconstruction of high-frequency information in the image. The super-resolution reconstruction of

Images