A C++ Implementation of Fine-to-Coarse Algorithm for Histogram-based Image Segmentation

Here we provide the source code and images used in the paper of "A C++ Implementation of Fine-to-Coarse Algorithm for Histogram-based Image Segmentation". It was written by Jin Huang and Julie Delon.

Requirements

Building and using requires the following libraries

libpng 1.2.54

The versions match the configuration we have tested on an ubuntu 16.04 system.

Compile

To compile the codes in Linux,

cd ./codes
make -j8

Usage

After compiling, you can input './ftc -h' to see the usage. usage : ftc input.png [eps] [-t] [-s sigma width] ftc [-h|-v]

     input.png   the name of input image
                 if image has 1 channel,
                 we will treat it as a gray image
                 if image has 3 channel,
                 we will treat it as a color image

     eps     value of epsilon
             default is 1
     -t      -t for using the third step
             the third step of ftc algorithm
             will cost lots of time
     -s      -s for using smooth
     sigma   the sigma used in smooth
     width   the bandwidth used in smooth

     -h      usage help
     -v      version info

Sample

./ftc ./images/flowers.png -s 3 3

Firstly, the program will output an image named "ftc_image_segmentation.png". It is the result of image segmentation. Secondly, there will be a txt file named "histogram.txt". This is the list of the histogram. Lastly, there will be a txt file named "index.txt", which is the list of the minimums ftc algorithm found.