The Natter toolbox is a Python toolbox for natural image statistics. It provides a unified framework for model comparison of natural image statistic models and for working with natural image patches in general. To evaluate the performance of natural image statistic models, comparability is crucial. The Natter toolbox provides a set of key features that simplify comparability through explicit logging of all data processing steps and use of reliable standard methods provided by python modules like NumPy, SciPy, MDP. The key features are for example:

The current release, version 0.2, can be downloaded here:
The complete documentation with several examples is also available online at The data used in the examples can be found here.
The latest version is available at our git repository You can check out the latest version with

git clone
If you find a bug or have ideas for improvement please use the issue tracker on the bitbucket website.


The functionality of the Natter natural image statistics toolbox is described in the following publication:
F Sinz, J-P Lies, S Gerwinn, M Bethge: NATTER: A Python Natural Image Statistics Toolbox. J Stat Software, 61,5. 2014
Please cite this paper if you are using the toolbox.


If you have any questions or problems, please contact natter AT bethgelab DOT org.

University of Tuebingen BCCN CIN MPI