Signal Dependent Local Noise Removal Using Weiner Filter Decomposition
P. Nagarathna, Afreen Kubra, G. Tirumala Vasu, Deepti Raj, Anitha Suresh, Samreen Fiza · Advances in computer science research · 2024
Image denoising finds applications in various fields like remote sensing, photography, biological imaging, astronomy etc.If image is corrupted with single source of noise, then a suitable denoising filter can be used.The major challenge associated with image denoising algorithms is denoising of image corrupted with multiple sources of the noise.Excessive smoothing can arise during the reduction of Additive White Gaussian Noise (AWGN) which can lead to a reduction in the level of detail and structural information and if Poisson noise is removed, then the AWGN components will still be retained in resultant image.To address this issue, we propose the Poisson Unbiased Risk Estimate Linear Expansion of Thresholds (PURE LET) approach that denoises mixed AWGN and Poisson noise images using Weiner filter decomposition.The application of a linear transformation to a filtered image allows for an inaccurate computation of the signal dependent local noise variance in the transform domain.Weiner filter inverts the blur of the image and removes extra noise by decomposing.The quantitative and qualitative analysis was conducted to determine the proposed algorithm's efficacy.