Filtering Techniques for Noise Reduction in Liver Ultrasound Images

Abstract

Ultrasound is an important diagnostic tool in diagnosing various abnormalities in the liver. Ultrasound is safe strategy for patient examination, being easy to apply, a non-invasive, and economical one. Ultrasound examination has the probability of repeatability producing images in real-time mode for diagnosing abdominal disorder such as liver cancer. Unprocessed ultrasound image has poor quality due to loss of texture or shape that can affect expert interpretation. The filtering methods are the initial process to reduce noise in the images. Enhanced image quality is needed for processing at the next stage after this initial processing. The research used salt and pepper, gaussian, and speckle noise. Filter that used to improved image quality to reduce noise are; mean, 2D medians, 3D medians, gaussian, and wiener. Image quality is measured by the quantitative value of MSE, RMSE and PSNR. Filter testing is done with noise level 1 and each filter has sigma value with average of 5. From the trial results, the best method for handling salt and pepper noise is the 3D median filter with MSE and RMSE values approaching 0 then having PSNR greater than 30 dB of all filter. The wiener filter is the best method to overcome Gaussian and speckle noise.