المجلة الدولية للعلوم والتقنية

Traditional image enhancement methods only consider the increase in brightness when enhancing low-exposure images, ignoring the problem of noise amplification during the process. Current deep learning-based methods, on the other hand, use end-to-end networks to learn the mapping relationship from low-exposure images to normal images without taking into account the physical principles of low-exposure image formation or the problem of noise amplification. In order to address the aforementioned issues, this paper examines the primary causes of image degradation and proposes a low-exposure image enhancement method based on a progressive dual network model. An image enhancement module and an image denoising module are both included in the method. Each module is built in a progressive manner, taking into account the image's brightness change from dark to light. This paper constructs a two-way constrained loss function to make the network learning result approach the real data from the positive and negative directions of the image degradation model to achieve dynamic balance, in order to better train the network. This article compares subjective and objective experiments with some mainstream methods in order to verify the efficacy of the method described in this article. The experimental results show that the results obtained using the method described in this article are more accurate and yield better performance indicators.