- This MATLAB code is an implementation for the segmentation algorithms proposed in:
Multi-Modality Feature Transform: An Interactive Image Segmentation Approach
M. Meshry, A. Taha and M. Torki
In British Machine Vision Conference (BMVC), 2015.

- For using this code, please cite the above paper.

- For computing the geodesic distance, we use the code provided by the Visual Geometry Group at Oxford University
http://www.robots.ox.ac.uk/~vgg/software/iseg/
The used function is called "perform_front_propagation_2d_color". To make it easy, we have compiled the required function to get the mex file for Linux 64-bit, Windows 64-bit and Mac 64-bit. The three versions are available in the "compiled mex files" directory. So, it is IMPORTANT to select the suitable file for your operating system and copy it to be beside the matlab files.
Alternatively, you can generate the mex file yourself according to your operating system. To do this, you will need to download the code from http://www.robots.ox.ac.uk/~vgg/software/iseg/ then run "compile_mex.m" found in "gsc-1.2/packages/+sp" 

- To run the code:
	- Make sure you have selected (or generated) the mex file corrosponding to your operating system, as illustrated above.
	- run "demo.m", which will run all 4 variations of our proposed approach
		1. 1-pass LDA
		2. iterative LDA
		3. 1-pass LFDA (O-MMFT)
		4. iterative LFDA (I-MMFT) --> best results

- Details of the approach are presented in the BMVC 2015 paper.

- For any problems/questions, you can contact the author at: moustafa.meshry@alexu.edu.eg
