The invention provides a super-resolution image
reconstruction method based on WPSN and CNN. The method comprises the following steps: constructing a training
data set by using LR image data and HR image data; constructing a new feature extractor
wave packet scattering network WPSN, and setting experimental parameters; constructing a new super-resolution image reconstruction model WPSN-CNN based on the WPSN, setting model hyper-parameters, training a
network model WPSN-CNN through the training
data set, and obtaining a trained WPSN-CNN
network model; and carrying out super-resolution image reconstruction by using the trained WPSN-CNN
network model. According to the method, the
wave packet system is introduced into the WSN, the
modulo operation in the WSN and the
convolution operation with the low-pass filter are removed, and the new feature extractor is established, so that the newly extracted features have translation exchangeability, and the multi-direction, multi-scale and multi-band features of the image can be fully utilized. The WPSN-CNN is used for performing a super-resolution image reconstruction task, so that richer image features can be extracted, a high-frequency structure and complex textures of an image can be reconstructed more accurately, the convergence speed on a small-scale
data set is higher, PSNR and SSIM indexes are remarkably improved, and the visual effect is more vivid.