Denoising of Multi Spectral Images Using Diagonal Tensor Decomposition Technique
Abstract
It's difficult to efficiently and effectively filter photos from more than one channel. Color and multispectral image (MSI) denoising combine contemporary nonlocal and transform domain algorithms to group similar patches to take use of natural images' self-similarity and sparse linear approximation. It is common to use a recursive procedure with a high number of comparable patches to model group level correlation in order to improve sparsity. A patch level representation's importance cannot be overstated. Our primary focus in this study is to examine the impact and possibility of representation at the patch level using a general formulation with a block diagonal matrix. We also demonstrate that a simple transform-threshold-inverse strategy can generate extremely competitive results by training a good global patch basis and a local principal component analysis transform in the grouping dimension. Fast implementation is also created to reduce computational complexity. Its robustness, effectiveness, and efficiency have been extensively tested on simulated and actual datasets.