Noise Reduction Using OBNLM Filter and Deep Learning for Polycystic Ovary Syndrome in Matlab

Noise Reduction Using OBNLM Filter and Deep Learning for Polycystic Ovary Syndrome in Matlab

Abstract

A hybrid speckle noise reduction method is proposed in this paper, which is a combination of OBNLM and deep CNN filters for better accuracy. Now-a-days polycystic ovary syndrome ultrasonic images be effecting through speckle noise. Speckle noise is the greatly affected noises in the ultrasonic images furthermore it is not easy to remove. In this paper, we are eliminating this speckle noise, which is the pre-processing step in image processing. The first step in pre-processing, which employs OBNLM (Optimized Bayesian Non-Local Means) filter and later CNN (Convolution Neural Networks). Here OBNLM filter is applied first and its output is given as input to CNN, generates a denoised image. The OBNLM provides a very strong filter for despeckling of Ultrasound images. CNN is an efficient method in deep learning, used also for noise reduction in ultrasound images. This work generates a denoising solution using the adaptive OBNLM filter in combination with the deep learning technique, which yields high performance. The performance of OBNLM, CNN and hybrid filter is measured using PSNR and SSIM.