tessa

A collection of Python modules for calculation of various texture features.

Features

Tessa can calculates features using these methods:

• SFTA (Segmentation-based Fractal Texture Analysis)
• CHOG (Circular Histogram of Gradients)
• RPST (Random Patch Dictionary with Soft Thresholding)
• TODO LPQ (Local Phase Quantization)
• TODO RZM (Regional Zernike Moments)
• TODO BGP (Binary Gabor Patterns)

Requirements

Tessa is written and tested for Python 2.7, and requires numpy >= 1.8 and scipy >= 0.13.3. The examples also depend on scikit-learn for classification.

Usage

TODO

Acknowledgments and citations

Christoph Gohlke's excellent tifffile code is included for reading and writing image data.

A subset of the Kylberg Texture Dataset is include for testing and demonstration purposes.

The SFTA module implements the algorithm presented by Alceu Costa in "An Efficient Algorithm for Fractal Analysis of Textures", SIBGRAPI 2012. It is based on his provided MATLAB source code

The CHOG module implements the algorithm published by Henrik Skibbe and Marco Reisert in *" Circular Fourier-HOG features for rotation invariant object detection in biomedical images", ISBI 2012. It is based on their published MATLAB source code

The RPST module implements methods described by Adam Coates and Andrew Ng in "The Importance of Encoding Versus Training with Sparse Coding and Vector Quantization", Proc. ICML 2011