An effective shearlet-based anisotropic diffusion technique for despeckling ultrasound medical images

The speckle is usually described as a multiplicative noise that significantly corrupts the quality and contrast in diagnostic ultrasound imaging. Therefore, the development of an effective noise reduction technique has become a crucial issue. Furthermore, the de-noising process should maintain the relevant medical information profitable for clinical diagnostic purposes. The goal of this paper is to propose an innovative technique for speckle noise removal. The core contribution of this work is the design of a new tensor-based anisotropic diffusion technique in the Shearlet domain. The idea consists of applying a directional smoothing process that depends on the local properties of the image that retains the essential structure details and improves the automatic stopping of the diffusion in the edge direction. Experimental results on both simulated and real ultrasound images indicate considerable enhancement in terms of PSNR, SSIM and ENL values carried out from the comparison of the proposed filter with existing standard denoising filters for speckle noise. These analysis results show that the proposed filter holds the higher value of PSNR (25.37 dB) and ENL (83.12) for simulated Phantom image. The experimental results on a group of representative filtered clinical data sets also demonstrate that STAD provides better smoothing performance with higher values of ENL (74.62) as compared to the state-of-the-art speckle filters. Hence in overall, these results reveal that the proposed model gives the best combination of speckle smoothing and edge retaining.