Image Denoising Using Wavelet Transform and Machine Learning

In image processing, blind picture deblurring is a difficult task that requires recovering a clear original image from a blurry and degraded version without knowing the blur kernel or the pristine image beforehand. In this paper, we present a novel method of improving the deblurring procedure by fusing two-dimensional discrete wavelet transform (DWT) with matrix-variable optimisation. By directly manipulating a matrix representation of the clean image, matrix-variable optimisation enables more accurate and efficient optimisation. This method works well for capturing the intricate spatial relationships present in the image, which makes it possible for the algorithm to navigate the complex structures found in blurry images. Additionally, the DWT decomposes the estimated clean image matrix into a number of frequency sub-bands. Better deblurring results are achieved by this decomposition, which not only makes the estimated image more regular but also helps to reduce high-frequency noise. Our method aims to overcome the difficulties caused by blind image deblurring by utilising DWT and matrix-variable optimisation in a synergistic way. Compared to conventional methods, our approach promises more precise and aesthetically pleasing results.