RESEARCH ON LOW-ILLUMINATION IMAGE ENHANCEMENT ALGORITHM

• 1、School of Electronic and Information Engineering, Liaoning Technical University, Huludao,Liaoning 125105

Abstract：At present, due to low light intensity and camera sensitivity limitation, many images are generated with low illumination, which seriously affects the subsequent processing and application of images. In this paper, a study aimed to improve the existing low-illumination image enhancement algorithms is carried out from the aspects of algorithm principle, enhancement effect and application scope.This study focuses on the integration of Convolutional Neural Networks (CNNs) into the RetinexNet algorithm. Through extensive simulation experiments, the impact of different parameters on the algorithm\'s performance is investigated, leading to proposed enhancements for the algorithm. The improvements primarily involve augmenting the training dataset of the RetinexNet algorithm and constructing a simplified network that comprises three components: enhancement, adjustment, and reconstruction. By combining Retinex theory and CNNs, this network facilitates automatic end-to-end processing and data-driven capabilities.our method not only achieves visually pleasing quality for low-light enhancement but also provides a good representation of image decomposition.

Keywords： Retinex theory Deep learning Image enhancement Low-illumination images