Radial B-Splines for Optimal Detection in Images
J. Fageot, V. Uhlmann, A. Depeursinge, Z. Püspöki, D. Sage, M. Unser
Proceedings of the Sixteenth IEEE International Symposium on Biomedical Imaging: From Nano to Macro (ISBI'19), Venice, Italian Republic, April 8-11, 2019, pp. 1849.
We present a two-step pipeline for the detection of patterns of interest in an image. An optimal detector is learned from a single template, and this detector is then used to retrieve the unknown location and orientation of all occurrences of the pattern in the image. Our approach relies on continuous-domain additive image model, where the analysed image is the sum of the template and a background signal with self-similar isotropic power-spectrum. We are then able to compute the optimal filter for the detection tasks, which maximizes the SNR criterion for this image model.
@INPROCEEDINGS(http://bigwww.epfl.ch/publications/fageot1901.html, AUTHOR="Fageot, J. and Uhlmann, V. and Depeursinge, A. and P{\"{u}}sp{\"{o}}ki, Z. and Sage, D. and Unser, M.", TITLE="Radial \mbox{{B}-Splines} for Optimal Detection in Images", BOOKTITLE="Proceedings of the Sixteenth IEEE International Symposium on Biomedical Imaging: From Nano to Macro ({ISBI'19})", YEAR="2019", editor="", volume="", series="", pages="1849", address="Venice, Italian Republic", month="April 8-11,", organization="", publisher="", note="")