Learning Iterative Binarization using Hierarchical Recurrent Networks
Author: Sven Behnke
In Proceedings of Joint 13th International Conference on Artificial
Neural Networks and 10th International Conference on Neural Information
Processing ICANN/ICONIP 2003, Istanbul, Turkey, pp. 306-309, June 2003.
In this paper the binarization of matrix codes is investigated as an
application of supervised learning of image processing tasks using a recurrent
version of the Neural Abstraction Pyramid.
The desired network output is computed using an adaptive thresholding
method for undegraded images. The network is trained to iteratively produce
the same output even when the contrast is lowered and typical noise is
added to the input. The network discovers the structure of the codes and
uses it for binarization. This makes the recognition of degraded matrix
codes possible for that adaptive thresholding fails.
Full paper: icann03.pdf
to selected image interpretation publications