Efficient analysis of DWT thresholding algorithm for medical image de-noising

This study proposes an efficient analysis based on objective and subjective test for filtering methods of the adaptive non-linear thresholding domain in discrete wavelet transform (DWT). The ultrasound images have been captured from three region-of-interests (ROIs), which are stomach, neck, and chest. These images have been converted into grayscale and three types of noises have been added, including Speckle, Gaussians, and Salt&Pepper. The objective test using Scilab 5 exhibits the performance of thresholding design on de-noising process in terms of signal-to-noise ratio (SNR). The visual effect has been measured using an inspection of image experts for de-noised image then yields the data of subjective test in terms of mean-opinion-score (MOS). In brief, this project succeeds to explore a new thresholding method with better performance in both SNR and visual effect named as hybrid estimated threshold (HET). It reveals that HET is the best filtering method to remove the Speckle, Gaussians, and Salt&Pepper noise.