Denoising of Medical Ultrasound Images Using Spatial Filtering and Multiscale Transforms

Medical imaging became the integral part in health care where all the critical diagnosis such as blocks in the veins, plaques in the carotid arteries, minute fractures in the bones, blood flow in the brain etc are carried out without opening the patient's body. There are various imaging modalities for different applications to observe the anatomical and physiological conditions of the patient. These modalities will introduce noise and artifacts during medical image acquisition. If the noise and artifacts are not minimised diagnosis will become difficult. One of the non-invasive modality widely used is ultrasound Imaging where no question of radiation but suffers from speckle noise produced by the small particles in the tissues who's size is less than the wavelength of the ultrasound. The presence of the speckle noise will cause the low contrast images because of this the low contrast lesions and tumours can't be detected in the diagnostic phase. So there is a strong need in developing the despeckling techniques to improve the quality of ultrasound images. Here in this paper we are presenting the denoising techniques for speckle reduction in ultrasound imaging. First we presented the various spatial filters and their suitability for reducing the speckle. Then we developed the denoising methods using multiscale transforms such as Discrete Wavelet Transform (DWT), Undecimated Discrete Wavelet Transform (UDWT), dual tree complex wavelet transform (DTCDWT) and Double density dual tree complex wavelet transform (DDDTCDWT). The performance of the filters was evaluated using various metrics based on pixel based, correlation based, edge based and Human visual system (HVS) based and we found that denoising using double density dual tree complex discrete wavelet transform is outperformed with best edge preserving feature.