Framework for noise-type and noise-level estimation under additive and multiplicative models in color images
Zipeng Fu, Xiaoling Ge, Xuelian Yu, Weixian Qian · Journal of the Optical Society of America A · 2025
Accurate estimation of the noise type and noise level in color images is crucial for tasks such as denoising, segmentation, and super-resolution. However, existing approaches often rely on the assumption that the noise type is known, and they tend to suffer from significant deviations when dealing with complex textures or strong inter-channel correlations in color images. To address these limitations, this paper proposes a quaternion-based framework for estimating noise type and noise level under two representative and widely used noise families: additive noise and multiplicative-additive noise (non-additive noise). By leveraging quaternion matrix modeling, the proposed method effectively captures cross-channel correlations, thereby enhancing the accuracy of both type discrimination between these two noise categories and noise-level estimation. On this basis, a classification model is developed by combining statistical features with logistic regression. Furthermore, differentiated noise-level estimation strategies based on weak-texture extraction are designed for the identified additive or non-additive noise models. Extensive experimental results demonstrate that the proposed method can accurately identify noise types and significantly improve the precision of noise-level estimation across diverse color image datasets and complex noise conditions, outperforming state-of-the-art techniques.