A Detail-Preserving Ultrasound Speckle Reduction Method Based on Complementary Information

Objective Speckle characterized by granular patterns resulting from wave interference in ultrasound (US) imaging systems, is considered an inherent common issue. The removal of speckle can greatly improve the clarity of the substructure in US images and enhance later processing. However, existing methods often involve a trade-off between denoising performance and detail preservation, which compromises their overall effectiveness. Methods To overcome this challenge, based on complementary information, this paper proposes a solution based on sparsity and second-order Hessian (SASH) regularization. Specifically, considering the inherent continuity of tissue structures in spatial distribution, we introduce continuity as a physical structural prior. By incorporating second-order Hessian derivatives to capture variations among neighboring pixels, the proposed method imposes constraints on the smoothness and directional consistency of tissue structures during the reconstruction process, thereby effectively suppressing speckle noise interference. Meanwhile, in order to counteract the loss of detail information caused by the denoising process, we further extend the concept of relatively sparse to enhance the high-frequency information which can better preserve the exquisite details of the image. Results By leveraging the complementary and synergistic effects of sparsity and structural priors, the SASH achieves a more effective compromise between noise reduction and detail retention. Conclusions SASH significantly enhances denoising and edge preservation capabilities.