Saak-Transform based Method for Image Noise Reduction

This research proposes a novel approach to reduce noise in an image. In this study, the input image is considered being shot with high ISO in the low light condition and the noise is modeled as the additive white Gaussian noise. The Subspace Approximation with Augmented Kernels (Saak) transform, a state-of-the-art spatial-spectral representation, is used for extracting the local characteristics of an image. The clean coefficients are estimated based on optimal linear minimum mean square error (LMMSE) estimation with a shrinkage on Saak coefficients. The processed image provides subjectively satisfactory quality improvement and an increase in peak signal to noise ratio (PSNR) without harming edges or other image details. It shows that Saak transform is a promising tool for noise reduction.