Neural network training method and device, image fusion method and device, equipment and medium

Abstract

The invention relates to the technical field of image processing and discloses a neural network training method and device, an image fusion method and device, equipment and a medium. The method comprises the following steps of designing a neural network comprising a first sub-network and a second sub-network with the same network structure, wherein any sub-network comprises a primary feature extraction module, a high-level feature extraction module and a coupling feedback module; the primary feature module is used for extracting low-level features of the under-exposure low-resolution image andthe overexposure low-resolution image; the high-level feature extraction module is used for further extracting high-level features from the low-level features corresponding to the underexposure low-resolution image and the overexposure low-resolution image; and the coupling feedback module is used for crosswise fusing low-level features and high-level features corresponding to the under-exposurelow-resolution image and the over-exposure low-resolution image. The method is advantaged in that multi-exposure fusion processing and super-resolution processing of the image are conducted at the same time through one neural network, the image processing speed is increased, and image processing accuracy is improved.

Description

technical field [0001] The present disclosure relates to the technical field of image processing, and in particular to a neural network training method, an image fusion method, a device, a device and a medium. Background technique [0002] With the development of technology, people are becoming more and more accustomed to using photos to record the details of their lives. However, due to the hardware limitation of the camera sensor, the captured images usually have various distortions, which make the images very different from real natural scenes. Compared with real scenes, images captured by cameras tend to have characteristics of low dynamic range (Low Dynamic Range, LDR) and low resolution (Low-Resolution, LR). In order to reduce the difference between the shot image and the real shot scene, the image needs to be processed. [0003] At present, multi-exposure image fusion (MEF) is mainly used to correct the low dynamic range problem of the image, and image super-resolut...

AI Technical Summary

Problems solved by technology

Moreover, the sequence of execution of the two image processing techniques sometimes affects the final image processing results

Therefore, the existing image processing methods not only have a cumbersome processing process, but also have unsatisfactory image processing effects.

Images