Speckle Filtering of Ultrasound Images Using Adaptive Total Variation Regularization in the Wavelet Domain

In recent years, ultrasound (US) imaging has been increasingly applied to the diagnosis of diseases. However, US images are usually degraded by speckle noise, which hinders the diagnostic process. To address this issue, many researchers have proposed various methods for speckle reduction. Nevertheless, these methods often fail to preserve fine structural details while removing speckle noise. In this paper, an improved despeckling method in the wavelet domain is proposed to achieve a balance between speckle reduction and detail preservation. Our approach employs speckle reducing anisotropic diffusion (SRAD) and total variation (TV) regularization to process the subbands, and introduces an adaptive regularization parameter estimation to the TV model. The proposed estimation method is based on the Markov random field (MRF) prior model and utilizes a hyperparameter to estimate the proportion of detailed information. Additionally, the regularization intensity of the TV model is adaptively adjusted according to the hyperparameter. Experimental results suggest that the proposed method not only effectively reduces speckle noise but also preserves critical details, thereby offering improved performance.