Non-local Hessian-based weighted filter for 2D noisy angiogram image enhancement with cardiovascular tree structure preservation

摘要

The Hessian-based filter had been already proven to be an effective method for enhancing vessels in 2D angiographic images. However, conventional Hessian-based filters are usually sensitive to noise and cannot preserve vascular structure very well. In this paper, a novel Hessian-based filter is proposed for enhancing 2D noisy angiography images. The proposed approach firstly incorporates a non-local operator in the Hessian matrix to obtain noiseless eigenvalues in a noisy environment, and then the relation of local image is carried by an exponent function which is used to suppress non-vascular structure around real vascular structure. Both qualitative and quantitative results of simulated experiments demonstrate that such approach is far superior to other two Hessian-based approaches in denoising, suppressing non-vascular areas, and enhancing vascular structure. Simultaneously, the results of extensive experiments on clinical X-ray angiography images further verify that our method performs better than the other two ones on obtaining high-quality enhanced vascular images which are beneficial for applications such as vessel segment, vessel motion analysis, and three-dementional reconstruction of vessel tree.