基于全局鲁棒Retinex模型的多通道弱光图像增强OA北大核心

In response to the issues where existing illumination enhancement methods ignore the changes in hue and saturation under different lighting conditions,resulting in the loss of excessive detail information while suppressing noise,we propose a global robust Retinex model(GRM)based on the Retinex theory to enhance illumination across all color channels.A contraction mapping is used to implement fractional-order norms,making the decomposed illumination map smoother.For the reflection map gradient,we employ a weight matrix and regularization term calculated based on the input image information to obtain a reflection map with less noise and more details.The alternating direction method of multipliers is chosen to solve the model,and gamma correction is applied to obtain the ideal illumination map.The processed reflection map is then combined with the illumination map to yield the final result.Experimental results on the Adobe FiveK,DICM,VV,and LIME datasets show that GRM not only competes with methods such as the robust Retinex model,low-rank regularization Retinex model,and plug-and-play Retinex model in terms of peak signal-to-noise ratio,structural similarity,and image quality assessment metrics,but also demonstrates stronger denoising capabilities and more significant image lighting recovery.The GRM model provides a superior solution for enhancing noisy images captured in low-light environments.