AI 中文总结
研究针对盲 RAW 去噪中 PG 参数估计受污染问题，提出 RPG-VST 框架，通过学生 t 对数方差目标为各 CFA 平面估计参数，用瓦片方差比估计$\sigma_z$作可靠性信号，在多数据集和去噪器设置中提升平均 PSNR 并减少严重尾部情况。
AI 中文摘要
使用广义安斯库姆变换（GAT）的方差稳定化使固定高斯去噪器能够处理泊松-高斯（PG）RAW 噪声，但其可靠性取决于拟合的散粒/读取噪声参数。在盲单图像部署中，这些参数从低纹理 RAW 统计中估计，而这些统计通常会被残留纹理、裁剪、缺陷像素和读取噪声底限破坏。这种污染会产生重尾对数方差残差，使普通最小二乘 PG 校准变得脆弱，并导致严重的尾部失败，尽管平均 PSNR 良好。我们提出了 RPG-VST，一种用于盲 RAW 去噪的鲁棒无参考方差稳定化框架。RPG-VST 使用具有鲁棒瓦片统计和物理约束的学生 t 对数方差目标，为每个彩色滤光片阵列（CFA）平面分别估计 PG 参数。然后，它从瓦片方差比中估计稳定域噪声水平$\sigma_z$，并将其用作可靠性信号。对于每个图像，RPG-VST 根据产生的$\sigma_z$更接近单位方差来选择鲁棒拟合或传统 OLS 拟合，无需干净参考或学习阈值。在 SID 索尼 SID50、SIDD 和 ELD 上使用固定的 SwinIR 和 Restormer 去噪器，RPG-VST 在所有六个数据集-骨干设置中提高了平均 PSNR。它在四种设置中减少了严重尾部，定义为 PSNR 增益超过直接方法低于-1dB 的情况，在其他两种设置中保持不变。在 SIDD 上，它产生+1.83/+1.92dB，并将严重尾部从 44/36 减少到 7/4。消融实验表明，$\sigma_z$门可防止在读取噪声主导的 ELD 捕获上无门控鲁棒拟合的回归。
英文摘要
Variance stabilization with the generalized Anscombe transform (GAT) enables frozen Gaussian denoisers to process Poisson--Gaussian (PG) RAW noise, but its reliability depends on fitted shot/read-noise parameters. In blind single-image deployment, these parameters are estimated from low-texture RAW statistics that are often corrupted by residual texture, clipping, defective pixels, and read-noise floors. Such contamination yields heavy-tailed log-variance residuals, making ordinary least-squares PG calibration brittle and causing severe tail failures despite favorable average PSNR. We propose RPG-VST, a robust no-reference variance-stabilization framework for blind RAW denoising. RPG-VST estimates PG parameters separately for each color filter array (CFA) plane using a Student-$t$ log-variance objective with robust tile statistics and physical constraints. It then estimates the stabilized-domain noise level $σ_z$ from tile variance ratios and uses it as a reliability signal. For each image, RPG-VST selects the robust fit or the conventional OLS fit according to which produces $σ_z$ closer to unit variance, requiring no clean reference or learned threshold. On SID Sony SID$50$, SIDD, and ELD with frozen SwinIR and Restormer denoisers, RPG-VST improves mean PSNR in all six dataset--backbone settings. It reduces severe tails, defined as cases whose PSNR gain over Direct is below $-1$ dB, in four settings and leaves them unchanged in the other two. On SIDD, it yields $+1.83/+1.92$ dB and reduces severe tails from $44/36$ to $7/4$. Ablations show that the $σ_z$ gate prevents regressions of ungated robust fitting on read-noise-dominated ELD captures.
CommentsAccepted for publication in IEEE Signal Processing Letters. 5 pages, 1 figure
