Speckle Noise Reduction for Ultrasound Images via Adaptive Neighborhood Accumulated Multi-scale Products Thresholding

摘要

Ultrasound imaging is widely used in medical diagnostic, unfortunately, the qualities of ultrasound images are generally limited due to the existence of speckle noises. As a result, edge-preserving noise reduction is an essential operation in ultrasound images processing. In this paper, we present an adaptive thresholding algorithm for ultrasound speckle suppression, which is based on dyadic wavelet transform (DWT) and neighborhood accumulated multi-scale products. Considering the dependencies between wavelet coefficients inter-scales, we multiply the adjacent sub-bands to intensify the edge and details while suppressing noise. Meanwhile, the probability of a large wavelet coefficient appearing in certain large wavelet coefficient's neighbors is great. We bring in the idea of neighborhood accumulated multi-scale products to exploit the intra-scale dependencies. The detail edges through our method can be more effectively distinguished from noise. Experiments show that the proposed method suppresses noise and preserves edges better than the state-of-the-art techniques.