Title: A combined hierarchical algorithm of mammograms registration using mutual information and a point based matching approach

Authors: Meryem Loucif; Bornia Tighuiouart

Addresses: Department of Computer Science, LRI Laboratory, Badji Mokhtar, Annaba University, 23000, Algeria ' Department of Computer Science, LRI Laboratory, Badji Mokhtar, Annaba University, 23000, Algeria

Abstract: Mammographic image registration is an important process in the comparative analysis of mammograms that aims to correctly interpret them. In this paper, we proposed a combined hierarchical framework for a non-rigid mammograms registration. So, the conjoint use of the multi-resolution and the multigrid hierarchical strategies allows to increase progressively the warp complexity with a fixe size of the manipulated data contributed then to a finest registration with acceptable computing cost. To perform registration process, a hybridisation of an intensity-based registration method using mutual information and a point-based matching approach using the Thin Plate Spline approach is achieved. This hybridisation allows not only to outperform the registration accuracy by using the mutual information as a similarity measure adapted to not similar images but also to decrease the algorithm convergence time. Registration performance has been evaluated by aligning mammograms from the MIAS database and results demonstrate the effectiveness of the proposed algorithm for mammograms registration.

Keywords: mammograms registration; MIAS database; mutual information; geometric matching; multi-resolution approach; Gaussian pyramid; combined hierarchical algorithm; progressive subdivision strategy; Thin Plate Spline.

DOI: 10.1504/IJBET.2019.097627

International Journal of Biomedical Engineering and Technology, 2019 Vol.29 No.3, pp.293 - 307

Available online: 18 Jan 2019 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article