Title :
Doubly-rotated matched filtering
Author :
Muhlich, Matthias ; Dahmen, Thorsten ; Aach, Til
Author_Institution :
Inst. of Imaging & Comput. Vision, RWTH Aachen Univ., Aachen, Germany
Abstract :
Using the concept of steerable filters, it is possible to represent the rotation of a known template - and thus arbitrary orientations of lines or edges - as a linear combination of a fixed number of base filters, thus allowing a computationally efficient detection of 1D structures. In spite of their usefulness in many image processing areas, such structures exhibit one important drawback: Due to the aperture problem, it is impossible to localize such features in images exactly. Junctions or corners, on the contrary, allow to find exact point matches in image sequences, but existing steerable filter approaches are not capable of modeling such templates. In this paper, we show how to extend rotated matched filtering using steerable filters such that it can represent multi-oriented image structures. As an example, we examine one special type of double-oriented feature which is especially important for applications like registration or camera calibration: the checkerboard pattern.
Keywords :
calibration; cameras; image registration; image sequences; matched filters; 1D structures; arbitrary orientations; base filters; camera calibration; checkerboard pattern; computationally efficient detection; double-oriented feature; doubly-rotated matched filtering; fixed number; image processing areas; image sequences; linear combination; multioriented image structures; registration calibration; steerable filters; Cameras; Europe; Feature extraction; Image edge detection; Interpolation; Signal processing;
Conference_Titel :
Signal Processing Conference, 2007 15th European
Conference_Location :
Poznan
Print_ISBN :
978-839-2134-04-6