Speckle Noise Removal Using Spatial and Transform Domain Filters in Ultrasound Images

Noise in the Ultrasound (US) images creates difficulties in interpreting the actual information in the image. Various studies have been done in this direction to retain useful information from the images. Speckle noise is an important noise type present in ultrasound images in the area of Synthetic Aperture Radar (SAR), active Radar, Optical coherence tomography (OCT) and medicine. The granular pattern of speckle noise is formed due to the simultaneous processing of backscattered signals from dispersed targets. This noise limits the quality and contrast of the obtained images resulting in poor understanding of the underlying facts. This work focuses on the various denoising techniques in spatial and transform domain for speckle noise removal in ultrasound images. The performance evaluation of these techniques was done with metrics like Speckle suppression index (SSI), Peak signal-to-noise ratio (PSNR) and Structural similarity index measure (SSIM). From the experiments, it is evident that SRAD algorithm in spatial domain really out performed all other filters with better SSI, PSNR and SSIM values.

I. Introduction

Denoising Ultrasound (US) images is one of the most demanding fields in image processing. Different noises like Gaussian, impulse, speckle are intrinsically present in US images, which corrupts the content of the image [1]. Among these, removal of speckle noise is challenging due to its multiplicative nature and estimating the noise distribution unlike Gaussian noise. Denoised medical images helps in increasing the certainty of the disease diagnosis and easily detect the lesions. SAR images that give high resolution pictures of the earth are also contaminated by speckle noise [2]. Eliminating speckle noise increase the accuracy of successive segmentation and classification tasks in these images. Therefore, despeckling is a major task in pre-processing stages of US images before detailed analysis and decision making. Ultrasound is high frequency sound waves, inaudible for humans which operates in the range of 2 MHz to 18 MHz [3]. In medical imaging, ultrasound transducers or probes send ultrasound waves to the body, which penetrates through tissues, and the device constructs the image from the echoed waves. Ultrasound waves is a series of small mechanical pressure waves propagating through the body. Ultrasound medical imaging is an effective, noninvasive and inexpensive imaging modality for obtaining cross section of internal organs [4]. The aim of this paper is to analyze the major denoising filters applicable to ultrasound images and quantitative evaluation between these techniques.