6 March 2018 Sparse representations via learned dictionaries for x-ray angiogram image denoising
Author Affiliations +
Proceedings Volume 10610, MIPPR 2017: Parallel Processing of Images and Optimization Techniques; and Medical Imaging; 106100C (2018) https://doi.org/10.1117/12.2285146
Event: Tenth International Symposium on Multispectral Image Processing and Pattern Recognition (MIPPR2017), 2017, Xiangyang, China
Abstract
X-ray angiogram image denoising is always an active research topic in the field of computer vision. In particular, the denoising performance of many existing methods had been greatly improved by the widely use of nonlocal similar patches. However, the only nonlocal self-similar (NSS) patch-based methods can be still be improved and extended. In this paper, we propose an image denoising model based on the sparsity of the NSS patches to obtain high denoising performance and high-quality image. In order to represent the sparsely NSS patches in every location of the image well and solve the image denoising model more efficiently, we obtain dictionaries as a global image prior by the K-SVD algorithm over the processing image; Then the single and effectively alternating directions method of multipliers (ADMM) method is used to solve the image denoising model. The results of widely synthetic experiments demonstrate that, owing to learned dictionaries by K-SVD algorithm, a sparsely augmented lagrangian image denoising (SALID) model, which perform effectively, obtains a state-of-the-art denoising performance and better high-quality images. Moreover, we also give some denoising results of clinical X-ray angiogram images.
© (2018) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Jingfan Shang, Zhenghua Huang, Qian Li, Tianxu Zhang, "Sparse representations via learned dictionaries for x-ray angiogram image denoising", Proc. SPIE 10610, MIPPR 2017: Parallel Processing of Images and Optimization Techniques; and Medical Imaging, 106100C (6 March 2018); doi: 10.1117/12.2285146; https://doi.org/10.1117/12.2285146
PROCEEDINGS
8 PAGES


SHARE
Back to Top