An end-to-end anti-blocking low-light image enhancement method

Abstract

The invention belongs to the field of image enhancement, in particular to an end-to-end anti-block effect low-illuminance image enhancement method. In the method of the invention, the collected data is first processed and then divided into two types: test data and training data; and then by constructing an end-to-end image enhancement network structure, the training data is input into the end-to-end network structure for multiple rounds of training. During the training process, the mean square two-norm difference between the network prediction map and the true value map is minimized, and the boundary loss of the compressed image is calculated to smooth the block boundary; and the block similarity loss value of the network prediction map and the true value map is calculated, Construct the structure similarity loss value to adjust the parameters of the end-to-end network, optimize the network structure, and obtain the end-to-end image enhancement network that has completed the training; then input the test data into the trained end-to-end image enhancement network structure to obtain the final enhancement image. Compared with the prior art, the method provided by the invention solves the problem of amplified image block effect of lossy compression and low illumination.

Description

technical field [0001] The invention belongs to the field of image enhancement, in particular to an end-to-end anti-block effect low-illuminance image enhancement method. Background technique [0002] Good lighting is an important factor to ensure image quality. When an image is captured from a low-light environment or the image obtained is an underexposed image, there are generally disadvantages such as low brightness, low contrast, and fuzzy details. It needs to be brightened. Enhanced processing for better visuals. Low-light image enhancement selectively highlights the features of interest in the image or suppresses some unwanted features in the image by transforming the data, so that the image matches the visual response characteristics, which can not only effectively improve the visual perception of images and videos, but also It provides a good preprocessing basis for vision processing applications such as target detection and tracking. In practical applications, due...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide an end-to-end anti-blocking effect low-light image enhancement method to solve the technical problem in the prior art that the blocking effect is amplified during the compressed image enhancement process

Images