Feature Preserving Regularized Savitzky–Golay Filter for Ultrasonic Images

Abstract

The speckle noise being multiplicative in nature degrades the quality of ultrasonic images. This noise limits the contrast resolution, complicates image-based quantitative and visual measurements and diagnosis. Therefore in this work, a novel denoising and feature preserving regularized Savitzky–Golay filter is proposed. The proposed work combines smoothing attribute of Savitzky–Golay smoothing filter and feature preserving regularization characteristic of total generalized variation to denoise ultrasound images. The proposed filter is applied and tested on synthetic images, Field II kidney simulated images and real ultrasound images. The conventional filters, i.e., speckle reduction anisotropic diffusion, non-local means and Savitzky–Golay smoothing filter are also designed for comparative analysis. The quantitative and qualitative comparison of results is carried out on the basis of PSNR, MSE, SSIM and Canny’s edge detection. It is revealed that the proposed filter denoises ultrasound images and preserves features more effectively in comparison with conventional methods.