Read the full-text of this article for free

Title: Line integral convolution-based non-local structure tensor

Authors: Yuhui Zheng; Kai Ma; Shunfeng Wang; Jing Sun; Jianwei Zhang

Addresses: School of Computer and Software, Nanjing University of Information Science and Technology, No. 219, Ning Liu Road, Pukou District, 210044 Nanjing, China ' School of Computer and Software, Nanjing University of Information Science and Technology, No. 219, Ning Liu Road, Pukou District, 210044 Nanjing, China ' Binjiang College, Nanjing University of Information Science and Technology, No. 219, Ning Liu Road, Pukou District, 210044 Nanjing, China ' Binjiang College, Nanjing University of Information Science and Technology, No. 219, Ning Liu Road, Pukou District, 210044 Nanjing, China ' College of Math and Statistics, Nanjing University of Information Science and Technology, No. 219, Ning Liu Road, Pukou District, 210044 Nanjing, China

Abstract: The non-local structure tensors have received much attention recently. However, the current computation methods of non-local structure tensor fail to fully use the anisotropic characteristic of tensors, hence resulting in limited performance. To address this problem, we present a novel anisotropic non-local regularisation scheme that integrates the atomic decomposition strategy with an extended line integral convolution method using non-local means filtering technique, in order to sufficiently utilise the spatial direction relevancy of tensors for their anisotropic smoothing. Experimental results on the test images show that our proposed anisotropic non-local structure tensor is superior to the current representative nonlinear structure tensors in corner detection.

Keywords: non-local structure tensor; image structure analysis; tensor field regularisation.

DOI: 10.1504/IJCSE.2018.089601

International Journal of Computational Science and Engineering, 2018 Vol.16 No.1, pp.98 - 105

Received: 09 Jun 2017
Accepted: 18 Aug 2017
Published online: 31 Jan 2018 *