基于饱和度–明度全变差的泊松噪声去除方法研究

A Study on Poisson Noise Removal Method Based on Saturation-Value Total Variation

摘要: 本文提出了一种基于饱和度–明度全变差(SVTV)的泊松噪声去除方法。该方法旨在克服传统泊松噪声保真项非线性问题的限制，提高彩色图像复原效果。首先，本文通过将图像转换至HSV颜色空间，利用饱和度和明度的全变差约束来保持图像边缘和细节。其次，使用泰勒展开对泊松保真项进行近似，以简化计算复杂度。实验结果表明，该方法在去除泊松噪声方面效果显著，相比基于RGB空间的传统方法，其在图像保真度、噪声抑制和视觉效果上均有较大提升。本文结论表明，SVTV与泊松保真项的结合具有较好的去噪能力，能够有效平衡噪声抑制和图像细节的保留。

Abstract: This paper proposes a novel Poisson noise removal method based on Saturation-Value Total Variation (SVTV) aimed at enhancing color image restoration. The objective of this study is to address the limitations associated with the nonlinearity of the classical Poisson fidelity term, thereby improving image restoration performance in the presence of Poisson noise. Initially, images are transformed into the HSV color space to apply total variation constraints on saturation and value channels, thus preserving edges and details more effectively. Subsequently, the Poisson fidelity term is approximated using a Taylor series expansion to reduce computational complexity and facilitate efficient optimization. Experimental results demonstrate that the proposed method achieves significant performance improvements in denoising compared to traditional RGB-space-based approaches. It offers better fidelity, noise suppression, and visual quality, showing the advantages of combining SVTV with the Poisson fidelity term. Conclusively, the study indicates that the SVTV method effectively balances noise suppression and detail preservation, making it a promising approach for Poisson noise removal in color image restoration.