Speckle Noise Reduction in Ultrasound Images Using Context-based Adaptive Wavelet Thresholding

Abstract

In medical imaging, image denoising has become a very essential exercise all through the diagnosis. Compromise between the preservation of useful diagnostic information and noise suppression must be respected in medical images. One of the ultimate goals of an imaging modality is to provide the clinician with the best possible information needed to make an accurate diagnosis. Ultrasound images suffer from an intrinsic artifact called speckle. Speckle degrades spatial and contrast resolution and obscures the underlying anatomy. Thus, it seems sensible to reduce speckle artifacts before performing image analysis, provided the image that might distinguish one tissue from another is preserved. The main goal of this thesis was the development of novel methods for suppression of speckle in medical ultrasound images in the wavelet domain. We have adopted weighted variance for estimating the threshold and multiscale product scheme for thresholding the coefficients. To employ the wavelet interscale dependencies, the adjacent wavelet subbands are multiplied. Multiplying the adjacent wavelet scales would sharpen the important structures while reducing noise. In the multiscale products, edges can be efficiently discriminated from noise. Then, an adaptive threshold was calculated and imposed on the products, instead of on the wavelet coefficients, to identify important features. Fundamentally speckle noise is a signal-dependent one, and so weighted variance of each background pixel was taken into account while calculating threshold. Experiments show that the proposed scheme better suppresses noise and preserves edges than other wavelet-denoising methods. Experiments with synthetic and real ultrasound imagery show that the proposed method improves the signal-to-noise ratio and preserves edge clarity.