Image enhancement method and device, terminal equipment and computer readable medium

Abstract

The invention provides an image enhancement method and device. The method comprises the following steps: converting the high-dynamic-range image into a normal-illumination low-dynamic-range image anda low-illumination low-dynamic-range image; in a process of performing iterative training on the self-encoding network through the sample pair data set, inputting the low-illumination and low-dynamicrange image into a self-encoding network; outputting the predicted low-illumination enhanced image through the self-encoding network; evaluating the difference between the predicted low-illuminance enhanced image and the normal-illuminance low-dynamic range image by adopting a mean square error; inputting the difference into the self-encoding network; adjusting the parameter value of the self-encoding network by the self-encoding network according to the difference; obtaining a trained self-coding network until the difference reaches a minimum value in an iterative training process, finally receiving a low-illumination image through the trained self-coding network, and outputting an enhanced image corresponding to the low-illumination image through the trained self-coding network, therebyavoiding the severe distortion of the generated enhanced image.

Description

technical field [0001] The invention belongs to the technical field of computer vision and intelligent transportation, and in particular relates to an image enhancement method, device, terminal equipment and computer readable medium. Background technique [0002] Images taken in low-light environments often have underexposure, resulting in overall dark images and blurred visual effects, which greatly affects the extraction and analysis of image information. Image enhancement is a commonly used image processing technique, which can increase the contrast of the image, thereby improving the visual effect of the image. [0003] As a research hotspot in the field of computer vision, deep learning has been increasingly applied to enhance low-light images. However, in the prior art, the neural network model fused with down-sampling convolution and up-sampling convolution is mainly used to enhance low-light images, which is likely to cause loss of image information. The enhanced im...

AI Technical Summary

Problems solved by technology

However, in the prior art, the neural network model fused with down-sampling convolution and up-sampling convolution is mainly used to enhance low-light images, which is likely to cause loss of image information. The enhanced image obtained by this method is seriously distorted, which is not conducive to subsequent images. deal with

Images