Improved Despeckle Filtering Technique for Liver Cirrhosis US Images

Abstract

The liver ultrasonic (US) images suffer from inherent speckle noise, degrading the image quality; thereby, affecting human interpretation and also reducing the accuracy in computer-assisted diagnostic techniques. In this paper, we propose an improved despeckle filtering technique by combining, maximum homogeneity over a pixel neighborhood filtering (MHON) and speckle-reducing anisotropic diffusion filtering (SRAD) using a binary classifier map (BCM). The textural and fine details of residual image are denoised by SRAD filtering. The SRAD filter denoised image and MHON filter denoised image combines to attain BCM. A BCM is generated by examining the local coefficient of dispersion (CoD) assessed using a kernel with the one obtained from a specific region. The proposed method is evaluated on clinical US image set of liver, by assessing the quality factors PSNR, SNR, ISC, LMSE and EPI. Results are compared with specified algorithms separately while accomplishing for proposed method.