Multi-supervision image super-resolution reconstruction method based on generative adversarial network

Abstract

The invention discloses a multi-supervision image super-resolution reconstruction method based on a generative adversarial network. The method comprises the following steps: registering slice images with different resolutions of the same slice; making a training data set by using the registered slice images; on the training data set, training a generative adversarial model by using a multi-supervision multi-stage generation model; and reconstructing the low-resolution image into a high-resolution image by using the trained generative adversarial model. According to the method, the image shot under the low-power lens is reconstructed into the high-resolution image, the imaging time can be saved, the hardware space for storing the image can also be saved, and blurring and artifacts of commonmethods on pathological data are overcome.

Description

technical field [0001] The invention belongs to the technical field of image processing, and more specifically relates to a multi-supervised image super-resolution reconstruction method based on a generative confrontation network, which is especially suitable for pathological slice images. Background technique [0002] At present, cell microscopic images require an objective lens with a magnification of 20x and above to be able to image clearly, but the imaging takes a long time and requires a large storage space. The imaging speed under the 4x objective lens is fast, and the requirements for equipment accuracy are low. However, the imaging resolution under the 4x objective lens is low, the depth of field is large, and the imaging is not practical. If images taken under a low magnification lens can be reconstructed into high-resolution images, both imaging time and hardware space for storing images can be saved. [0003] Some researchers at home and abroad have proposed th...

AI Technical Summary

Problems solved by technology

In recent years, super-resolution methods based on deep learning have achieved remarkable results, but most of the models are based on degraded data, and there are certain limitations in the application of images of different resolutions collected from the real world.

[0004] In summary, low-magnification objective lens imaging super-resolution has great practicability, but various super-resolution reconstruction methods have limited effect on pathological data

Images