Local Statistical-Based Filtering for Different Noise Models: A Comparison

This paper shows the performance of 6 local statistics based filters for 4 different types of noise models. Most of the research articles preferred only one type of noise especially speckle noise removal filters hence, this focus on speckle, Gaussian, Poisson and salt & pepper noise for analyses of noise suppression, edge, and structure preservation evaluated in terms of image quality metrics, visual quality assessment. The comparative analysis reveals that wiener filter has better Signal to Noise Ratio (SNR), Figure of Merit (FOM) and Structural similarity index (SSIM) for speckle and Poisson noise images and lsmv filter has better Image quality index (IQI) and Beta metric (β) image quality metrics for speckle noise images and no filter has all better image quality metrics for all types of noise.