Ultrasound B-scan image speckle reduction and super-resolution

Date of Completion

January 2004


Engineering, Biomedical|Engineering, Electronics and Electrical




This dissertation work has been focused on the important issues of ultrasound B-scan image speckle reduction and super-resolution reconstruction. As an important element of the research, we dealt with the speckle modeling issue first. Starting from the statistical analysis of the dynamic range uncompressed speckle field, we derived the statistical models for the dynamic range compressed situation. Based on the statistical analysis, we provided some important guidelines for further development. ^ Instead of considering an ultrasound B-scan image as a set of points, it can be viewed as a set of line elements. In structural regions, the elements are more directionally clustered than in the homogeneous regions. We designed directional cancellation masks to identify these elements and perform smoothing along them to achieve speckle reduction while preserving the edge information. ^ Using a different approach, we applied the anisoptropic diffusion scheme to speckle reduction. The diffusion technique can be designed to enhance the actual edges instead of simply preserving them. To assure a stable solution, we applied the median regularization into the diffusion process. The image decimation is used to improve the efficiency of speckle reduction and the computational cost. The experimental data have shown the success of the above two proposed techniques. ^ Super-resolution reconstruction is a high-level image restoration technique. It recovers a high-resolution image from a set of sub-pixel shifted low-resolution images. In our work, a non-homogeneous anisotropic diffusion process is proposed. This method recovers the high-resolution details and enhances the structure information effectively while removing noise. This is evidenced by the experimental results. ^ Finally, we proposed a pixel-compounding technique to reconstruct super-resolved ultrasound B-scan images. First, the point spread function of the imaging system is estimated by homomorphic transformation, then the images are restored at low-resolution level, and finally the anisotopic diffusion super-resolution reconstruction technique is applied to recover a high-resolution image. A phantom study showed 100% resolution improvement in terms of the average half-peak-width (AHPW) metric. This technique has the potential to improve the measurement accuracy of the carotid artery intima-media thickness, which is an important biomarker for prognosis and diagnosis of atherosclerosis, a peripheral circulation disease, and potential stroke. ^