Search papers, labs, and topics across Lattice.
This paper introduces a fidelity-constrained framework that effectively anchors the output of black-box denoisers to their input without the need for retraining, utilizing a linear blending approach. By selecting the maximum blending factor that adheres to local fidelity constraints defined by Peak Signal-to-Noise Ratio (PSNR) or Structural Similarity Index (SSIM), the method achieves a balance between denoising performance and statistical naturalness. Experimental results demonstrate that SSIM-based anchoring outperforms PSNR-based methods in maintaining consistency across varying noise levels while effectively controlling fidelity.
SSIM-based anchoring offers a more reliable fidelity control for denoising outputs, outperforming traditional PSNR methods across diverse noise conditions.
We propose a fidelity-constrained framework that anchors the output of a black-box denoiser to its input without retraining and with little additional computation. The method linearly blends the denoised image with the input and selects the maximum blending factor that satisfies a prescribed local fidelity constraint using Peak Signal-to-Noise Ratio (PSNR) or Structural Similarity Index (SSIM). For PSNR control, a closed-form solution is obtained under a local constant-blending assumption. For SSIM control, we derive a tractable formulation based on inverse SSIM under the same assumption and solve it efficiently using iterative root finding. Experiments on DIV2K images with synthetic Gaussian noise and outputs from Real-ESRGAN and a non-local means denoiser show that the proposed anchoring strategy provides effective fidelity control while balancing denoising performance and statistical naturalness, as measured by the excess kurtosis of residual noise. In particular, SSIM-based anchoring yields more consistent behavior across noise levels than PSNR-based anchoring.
