by Sebastian Zambanini
Abstract:
This report presents an automatic method for registering follow-up hemangioma images taken during clinical trials in specific time intervals. The method finds interest points in two images on the basis of edge points and matches corresponding interest points using SIFT features. Under the assumption that hemangioma regions are planar, these correspondences are used to determine a homography between the two images by means of RANSAC. Experimental results are reported for image pairs acquired at the same time and for image pairs acquired during follow-up showing hemangiomas at different times. Registration errors on the images are acceptable for subsequent processing, however, gross changes of hemangiomas deteriorate the performance and are subject to ongoing research.
Reference:
Automatic Registration of Cutaneous Hemangiomas in Digital Image Series (Sebastian Zambanini), Technical report, PRIP, TU Wien, 2007.
Bibtex Entry:
@TechReport{TR114,
author = "Sebastian Zambanini",
title = "Automatic Registration of Cutaneous Hemangiomas in Digital Image Series",
institution = "PRIP, TU Wien",
number = "PRIP-TR-114",
year = "2007",
url = "https://www.prip.tuwien.ac.at/pripfiles/trs/tr114.pdf",
abstract = "This report presents an automatic method for registering follow-up hemangioma images taken during clinical trials in specific time intervals. The method finds interest points in two images on the basis of edge points and matches corresponding interest points using SIFT features. Under the assumption that hemangioma regions are planar, these correspondences are used to determine a homography between the two images by means of RANSAC. Experimental results are reported for image pairs acquired at the same time and for image pairs acquired during follow-up showing hemangiomas at different times. Registration errors on the images are acceptable for subsequent processing, however, gross changes of hemangiomas deteriorate the performance and are subject to ongoing research.",
}