Medical image super-resolution reconstruction method based on network cascading

Abstract

A medical image super-resolution reconstruction method based on network cascading comprises the following steps that: 1) an image is preprocessed, so that the image can better participate in training,and learning of a subsequent network is facilitated; 2) an image pyramid is constructed, and cascade information of the images are extracted; 3) super-division is carried out on the image by utilizing a cascade network; 4) the design basis of the cascade network is a Laplace pyramid and a BE-GAN, and the network generates a high-definition super-resolution image by generating a residual image andadding the residual image to an original image; the total amount of information required by network learning can be reduced by using the residual image, and the training pressure of the network can be reduced, so that the network can obtain a better effect under a smaller data set; cascading enables all levels of the network to be relatively independent, so that the network can adapt to super-division tasks under all resolutions.

Description

technical field [0001] The invention relates to a medical image super-resolution reconstruction method. Background technique [0002] At present, medical imaging has become an important diagnostic basis and tool in clinical medicine, participating in the diagnosis and treatment of diseases. With the continuous development of modern medical imaging technology, there is a significant difference in clarity between images taken in the past and those taken today. In order to better conduct research on diseases and follow up patients, it is necessary to reduce this difference, which requires super-resolution processing of images. Common main methods are: bilinear interpolation, bicube interpolation, convex set projection method (POCS), maximum a posteriori probability estimation method (MAP), field embedding method (NE), sparse representation (SC) and neural network-based SRCNN, DCNN, SRGAN, BEGAN, etc. However, due to the algorithm design based on interpolation calculation in ...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the deficiencies of the prior art, the present invention combines the Laplacian pyramid model in iconography to provide a medical image super-resolution reconstruction method based on network cascading. Performance on small-sample datasets of medical images; using neural networks to solve the problem of image quality degradation during image super-resolution

Images