Abstract
We consider neural network classifiers which can estimate Bayesian a posteriori probabilities1 as Ideal Observers (IO) for classification of stochastic image patterns for which the Human Observer (HO) is known to be less reliable. Such situations arise across different imaging modalities of interest: medical X- and gamma rays and ultrasound; and passive and active imaging in the visible and the infrared.
© 1993 Optical Society of America
PDF ArticleMore Like This
Tin M. Aye, Araz Yacoubian, Dai-Hyun Kim, and Tomasz Jannson
ThI.3 OSA Annual Meeting (FIO) 1993
E. Kakkava, N. Borhani, Y. Pu, C. Moser, and D. Psaltis
Th5A.2 Conference on Lasers and Electro-Optics/Pacific Rim (CLEO/PR) 2018
Piergiorgio Caramazza, Roderick Murray-Smith, and Daniele Faccio
CW3B.4 Computational Optical Sensing and Imaging (COSI) 2018