A Study and Exploration of Discrete Wavelet Transform for Speckle Noise Reduction in Ultrasound Images
Paradee Namsopa, Santi Koonkarnkhai, Piya Kovintavewat, Harutai Dinsakul, Sopapun Suwansawang · 2023
This study aims to compare different discrete wavelet transform for speckle noise reduction in ultrasound images. A set of ultrasound images were filtered using Haar wavelet, different scaling of three basic wavelet bases, Duabechies and Symlets wavelet. We conducted an experiment by adjusting the value of the constant$k$by proposed thresholding rule. We used objective metrics to evaluate our analysis namely Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR), Structure Similarity Index (SSIM) and Coefficient of Correlation (COC). The experimental results demonstrate that the proposed thresholding rule leads to improved image performance compared to existing VisuShrink thresholding rule ($k=1$). In addition, the results show that the optimal value of$k$for all wavelet families used in this study is between 0.5 and 0.7. The sym8 provides the best performance compared to other wavelet types used in the experiment.