Noise Reduction using Multilevel Hybrid Filters for Ultrasound Images: A Comparative Study

摘要

Medical imaging has become a very important non-invasive diagnosis tool for various diseases. During image acquisition, they get acquainted with noise because of modality and/or because of body conditions such as patient’s position or body fat. Most common noise in ultrasound is speckle, CT and MRI are motion and electrical noises. And other additional noise can also be present such as salt and pepper, Gaussian, Poisson. These noises degrade image quality. The only expert radiologist can make an appropriate diagnosis. Hence it is very important to perform de-noising. Many techniques are present for de-noising which are basically dependent on noise type. Here an attempt is made to remove speckle noise using multilevel hybrid filters. Ultrasound images of the common carotid artery are degraded by speckle noise and then de-noised using the median, wiener, NLM, Homorphic, Bilateral and hybrid filter are applied. Performance of these algorithms is measured using PSNR, MSE, SSIM and ROC curve is drawn. Results show that the Median filter With Bilateral filter has better performance.