APILASTRI / Image-Segmentation-Metrics-

Object Oriented Error Measures project 01/2016

Released under the GPL license 3 http://www.gnu.org/licenses/ written by: Oscar Alonso Cuadros Linares [e-mail]

This project aims to implement object-oriented measures to quantify the quality of image segmentation algorithms. Currently, there are three implemented measures:

1. Alberlaez Error Measure (AEM), Arbelaez et. al. (2009)

2. Object-level Consistency Error (OCE), Polak et. al. (2009)

3. An Adjustable Error Measure for Image Segmentation Evaluation (AOM), Oscar Cuadros Linares et. al. (2015): This error measure outperforms both AEM and OCE measures, not only in terms of accuracy but also in time processing. Besides, AOM satisfies the three axioms of metric spaces.

Usage:

This project is implemented in C++, you only have to include the header file "metric.h" into your own project. Moreover, we implemented a class to read and write SEG files and a class to measure the processing time. See the example below:

#include "metric.h"

int main() {

SEG seg_1, seg_2;
seg_1.read("input/square_1.seg");
seg_2.read("input/square_2.seg");

AOM aom;
aom.penality(0.5);
std::cout << aom.error(seg_1, seg_2) << std::endl;

/*****************/

Metric *oce = new OCE();
Metric* arbelaez = new Arbelaez();

Test test;

test.chrono(oce, seg_1, seg_2);
test.chrono(arbelaez, seg_1, seg_2);

delete oce;
delete arbelaez;

return 0;

}

TODO: