A FCM-based image de-noising with spatial statistics pilot study

Abstract

Image de-noising is an important scheme to make the image visually prominent and obtain enough useful information. To obtain reliable results, many applications had developed for effective noise suppression and received good image quality. This report assumed a residual image consisted of noises with edges produced by subtracting the original and a low-pass filter smoothed image. The Moran statistics was then using to measure the variation of spatial information in residual images and then as a feature data input to FCM. Three clusters pre-assumed for FCM in this work: they are heavy, medium and less noisy areas. The rates of each position partially belongs to each cluster were determined by a FCM membership function. Each pixel in noisy image assumed in de-noising processing that which is a linear combination of product of three de-noised images with membership function in the same position. Average filters with different windows and a Gaussian filter priori applied to this noisy image to make three de-noised versions. The results showed that this scheme worked better than the non-adaptive smoothing. This scheme‘s performance is evaluated and compared to the Bilateral filter and NLM using PSNR and SSIM. The developed scheme is a pilot study on this area. Further future studies needed on the optimized number of clusters and smoother versions used in linear combination.