ISSN 2079-3537      

 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                                                                                             
Scientific Visualization
Issue Year: 2014
Quarter: 4
Volume: 6
Number: 4
Pages: 61-71
Article Name: FAST REGISTRATION ALGORITHMS FOR HISTOLOGICAL IMAGES
Authors: D.I. Sungatullina, A.S. Krylov, D.N. Fedorov (Russian Federation)
  The paper is recommended by program committee of 24th International Conference on Computer Graphics and Vision GraphiCon’2014.
Address: D.I. Sungatullina
diana.sungatullina@gmail.com
Laboratory of Mathematical Methods of Image Processing of the Faculty of Computational Mathematics and Cybernetics of Lomonosov Moscow State University, Russia

A.S. Krylov
kryl@cs.msu.ru
Laboratory of Mathematical Methods of Image Processing of the Faculty of Computational Mathematics and Cybernetics of Lomonosov Moscow State University, Russia

D.N. Fedorov
dnfedorov@mail.med.ru
Federal State Scientific Institution “B.V.Petrovsky National Research Center of Surgery”, Russia
Abstract: In this paper, we propose two efficient registration algorithms for histological images. Our first algorithm is based on matching the contour points of an observation and a template, and performs in O(M log M) time, where M is the number of the contour points, with almost no loss in the quality of registration compared to the commonly used Hungarian algorithm, which has O(M3) time complexity. A high accuracy of the transform parameter estimation is achieved by an iterative exclusion of heavily mismatched contour points, followed by rectification of the parameters for the rest of the points. Our second algorithm takes advantage of a special structure of the histological images that contain elliptical gland slices, and finds corresponding ellipses on the observation and the template. The resulting transform is obtained from the pair of ellipses with maximum overlap index. The first method suits well for all types of histological images, while the second one is intended to be used for high-precision content-based alignment.
Language: English