Low-illumination image enhancement method based on deep learning

Abstract

The embodiment of the invention discloses a low-illumination image enhancement method based on deep learning. The method comprises the following steps: step 1, constructing a multi-exposure and multi-scale reflectivity image extraction model; 2, constructing a brightness weight matrix estimation model, wherein the brightness weight matrix estimation model is used for estimating an input image brightness weight matrix and a reflectivity image brightness weight matrix; 3, constructing an illumination enhancement model, and combining the output of the multi-exposure and multi-scale reflectivity image extraction model and the brightness weight matrix estimation model to establish the illumination enhancement model for predicting an input low-illumination image to obtain a normal-illumination image; and 4, constructing a loss function, carrying out training based on a gradient descent method, and solving optimal parameters of the low-illumination image enhancement overall model. According to the embodiment of the invention, the problem that the image processing effects of recovering details and keeping colors are difficult to integrate in the existing low-illumination image enhancement technology is solved, so that the visual impression effect is improved.

Description

technical field [0001] The present invention relates to but not limited to the field of image processing, in particular to a low-light image enhancement method based on deep learning. Background technique [0002] In recent years, the research on low-light image enhancement technology is one of the hotspots in image processing. The implementation methods are divided into traditional image enhancement methods based on Retinex theory and new image enhancement methods based on neural networks. [0003] Traditional image enhancement methods based on Retinex theory, including MSR, MSRCR and MSRCP methods, etc. The above methods have obvious effects in low-light image enhancement, but they all have problems such as long processing time and color distortion, which make the visual perception effect not good. [0004] New image enhancement methods based on neural networks, including RetinexNet, LLNet, MBLLEN and KinDNet methods, etc. Most of the above methods have better processing...

Images