An Image Super-Resolution Reconstruction Method Based on Wavelet Coefficient Learning

Abstract

The invention discloses an image super-resolution reconstruction method based on wavelet coefficient learning, which belongs to the field of image super-resolution. In the present invention, image training set and test set are selected first; two-dimensional stationary wavelet transform is performed on low-resolution images, and two-dimensional wavelet packet transform is performed on high-resolution images to obtain wavelet coefficients of LR images and HR images respectively; step 3: constructing Deep neural network; Step 4: Use the deep neural network to extract the deep network features of the LR image wavelet coefficients, and obtain the HR image wavelet residual coefficients; Step 5, restore the obtained HR image wavelet residual coefficients to HR wavelet coefficients , and perform inverse two-dimensional wavelet packet transform to obtain the final high-resolution image. Compared with the prior art, the present invention can perform super-resolution reconstruction on images in a very short time by using a large number of external training sets to train a deep super-resolution model, and the reconstruction results are obviously better than most algorithms.

Description

technical field [0001] The invention belongs to the field of image super-resolution, in particular to an image super-resolution method based on wavelet coefficient learning. Background technique [0002] Single Image Super-Resolution (SISR) refers to a given single low-resolution image, through a specific algorithm to restore the image details lost due to downsampling, so as to obtain more detailed information. , The process of high pixel density images with more detailed picture quality. Due to its ability to recover finer details with limited information, this technology has been widely used in various application scenarios, such as high-definition TV playback, video surveillance, and satellite imaging. Since the downsampling process of a high-resolution image under different conditions will obtain a low-resolution image with different lost details, the reconstruction process from a low-resolution image to a high-resolution image is a typical ill-conditioned inverse probl...

AI Technical Summary

Problems solved by technology

2) For gradient disappearance / explosion, it is difficult to converge

[0005] However, the above CNN-based methods also have some shortcomings: 1) It is easy to produce over-smoothed (Over-Smoothed) results, lacking some detailed information; 2) Often use fixed upsampling operators, such as cubic spline interpolation operators, in Increasing the image resolution also introduces distortion information and increases the computational burden of the network; 3) It requires a lot of calculations and the training phase takes too long

Images