Shearlet and guided filter based despeckling method for medical ultrasound images

Abstract

Ultrasound diagnostic techniques are widely used in medical clinical diagnostics. However, the presence of speckle noise in the ultrasound imaging process reduces the image quality and make inconveniences for the clinical diagnosis of the physician. Therefore, to decline the influence of speckle noise has important significance in medical ultrasound image diagnosis. In order to solve the problem of speckle noise, this paper proposes a novel de-speckle method for medical ultrasound imaging, which is based on nonsubsampled shearlet transformation (NSST) and guided filter. Firstly, the nonsubsampled Laplacian pyramid filtering is used to decompose the noisy image, and thus the image is decomposed into high frequency subbands and low frequency subbands. Under the direction of non-sampling filter bank, we obtain a high-frequency subband multi-directional decomposition; Secondly, based on the threshold function and the correlation of the shearlet coefficients in the transformation domain, an improved threshold shrinkage algorithm is proposed to perform the threshold shrinkage processing on the shearlet coefficients of the high frequency subbands. Finally, the low frequency subbands in the transformation domain are processed by the guided filter, and the denoised ultrasonic image is obtained by the inverse transformation of the shearlet. In order to verify the effectiveness of the proposed method, experiments are conducted, and the results were compared with those of other existing denoising filter. Showing that the proposed method has a strong de-noising performances and maintain image details.