ywg121020 / ADMD

Absolute directional mean difference (ADMD) algorithm for small infrared target detection

ADMD

Absolute directional mean difference (ADMD) algorithm for small infrared target detection

Both MATLAB and OpenCV implementations of the following paper: Fast and robust small infrared target detection using absolute directional mean difference algorithm

If you are using these codes in your researches, kindly cite the following paper:

Moradi, Saed, Payman Moallem, and Mohamad Farzan Sabahi. "Fast and robust small infrared target detection using absolute directional mean difference algorithm." Signal Processing 177 (2020): 107727

MATLAB

You can find the single-scale (7by7) implementation of the ADMD algorithm in MATLAB subdirectory. The multi-scale version can be easily constructed by max selection among different scales (see the paper for further information). To achieve saliency-map, just pass the test image through AdMD7_eff function:

test_img=double(test_img);
Filtered_image = AdMD7_eff(test_img);

OpenCV

Both single and multi-scale implementation of the ADMD algorithm can be found in CPP subdirectory. The codes are implemented using OpenCV 3.4 in Ubuntu 18.04 environment.

Prerequisites

OpenCV

Setup

Build OpenCV from source (if you want to enable QT option to zoom on images and see the pixels value), or just simply use the following command in terminal:

$ sudo apt install libopencv-dev python3-opencv