Year: 2025
Type: article
Abstract: Images captured in dark environments often face challenges like low contrast and noise. Many enhancement methods based on variational models, set parameters artificially to improve contrast, which often does not adequately address real-world conditions. To address this, we propose an adaptive variational model and establish an Adaptive Variational Iteration Network (AVINet). By designing a Parameter Extraction Module, we can adaptively learn the image contrast adjustment parameters within the model. Then, to prevent uneven image brightness, we introduce a smoothness term. We have mathematically proven the existence and uniqueness of the solution of this variational model, and then solve it using the Bregman iteration method. In the Denoising Module, we employ a denoising diffusion probability model to eliminate the noise in the image. Comparative experiments conducted on various datasets demonstrate that that our method achieves better results in restoring illumination and detail compared to other advanced methods.
Authors: Tao Chen, Feng-Fei Jin, Jinqi Cui, Xiangyu Man, Dongmei Liu
Institution: Shandong Normal University
Language: English
Topic: Image Enhancement Techniques
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,990
Awards: G1273438263, G4397726962, G694022097
OpenAlex ID: W4413451349