Low-light image enhancement based on LAB color space

In recent years, the field of image enhancement has gained significant attention, especially for images taken in dimly lit settings. These images frequently exhibit issues such as diminished contrast, reduced brightness, and disruptive noise, all of which can detrimentally affect their quality. Given the swift progress of deep learning technology, a multitude of methods for enhancing low-light images have emerged, leveraging deep convolutional neural networks (CNNs). Through an understanding of the relationship between low-light images and standard images, deep convolutional neural networks can adeptly extract image characteristics and improve overall image quality. In this paper, we present a deep convolutional neural network-based method that uses a deep unsupervised dehazing network model as a foundation. Our method stands out from existing ones by considering both algorithmic decomposition and image decomposition. It achieves low-light image enhancement goals by breaking down the enhancement problem into three sub-problems through laboratory space decomposition of the image. Experiments show that our proposed method can lead to significant enhancement of images with low light intensity.