WND-CHARM - A multi-purpose image classifier.

The wndchrm command-line utility gets a root directory as a parameter and computes a large set (up to ~2700) image features for each image in the class subdirectories based on the WND-CHARM algorithm. It can then split the images into training and test sets and classify them. The code was tested on MacOS X and Linux. Use "configure" and "make" after untaring the files. Then type "wndchrm -h" for instructions.

Full description of wndchrm command line utility and the WND-CHARM algorithm can be found at:

Shamir, L., Orlov, N., Eckley, D.M., Macura, T., Johnston, J., Goldberg, I., Wndchrm - an open source utility for biological image analysis, BMC Source Code for Biology and Medicine, 3: 13, 2008.

Download wndchrm: wndchrm.tar.gz


Important:
Wndchrm requires the installation of two publicly available libraries - LibTIFF and FFTW. The libraries cab downloaded from:

LibTIFF: http://www.libtiff.org
FFTW 3.1: http://www.fftw.org/download.html


Systems requirements:
Recommended Hardware: 512 MB RAM, 10 MB HD space, 2 GHZ CPU. Software: GCC compiler must be installed. Wndchrm was tested using Linux (Fedora Core, Red Hat) and MacOS X, but should also run under Unix and other Linux distributions.
Installation difficulties are expected if Libtool is compiled with a different compiler version than the one used for compiling LibTIFF (this situation is more likely with upgraded operating systems). This problem can be solved by installing a newer gcc version.


MS-Windows users can download a binary executable file wndchrm.exe

The executable file needs the following run-time DLLs:
fftw3.dll
libfftw3-3.dll
libtiff3.dll
jpeg62.dll
zlib1.dll
libsndfile-1.dll

MS-Windows developers can download compiled libraries for Borland C++ and Microsoft Visual C++