Pyramid-based multiresolution adaptive filters for additive and multiplicative image noise

In this work, adaptive filtering is applied to the different resolutions of the Laplacian pyramid of a noisy image. For natural images, each layer is characterized by an SNR that increases as resolution decreases. Therefore, each filter may be adjusted to achieve adaptivity also across scales. Theoretical frameworks are developed for signal-dependent noise models. A rational Laplacian pyramid is employed to handle multiplicative noise. Experiments carried out on gray-level images corrupted by true and simulated noise, show the potentiality of the approach compared to other established algorithms, in terms both of SNR improvements and of enhancement in visual quality.