---------------------------
- plugins to_numpy and from_numpy added for support of numpy;
the deprecated numeric and numarry modules have been replaced with numpy
- highlight also works with GREYSCALE and ONEBIT images
- corrected resize function in VIGRA
- the knn classifier can now return different confidence
measures for the main id that are selectable by the
user. See the classifier API documentation for details.
- knn classifier now works as expected when k > 1;
up to now, id_name was sorted by confidence, which meant
that always the farthest among the k nearest neighbors
was returned as main_id
- now compiles with Python 2.6 (thanks to Prapat Suriyaphol)