Image-to-Region-Adjacency-Graph-creation

Paper | MATLAB

Convertion of an RGB image to a Region Adjacency Graph (RAG) using SLIC super-pixel based segmentation technique.

1. segmentation.m - contains the type of segmentation algorithm. The code by default uses SLIC superpixel based segmentation. However, a graph-cut based segmentation implementation can also be found in the commented section.
2. Extractfeaturevec.m - contains the features to be extracted from each nodes.
3. filterimage.m - This is the main file. We have used the single labelled UC-Merced dataset, containing 21 classes. Mention the path to these images in the "addpath" line. Totgra saves the final matrix containing the node features, wighted adjacency matrix information, etc.

#Requirements- Set up vlfeat library for using the SLIC super-pixel based segmentation.

Find the UC-Merced dtaset from http://bigearth.eu/datasets.html

Paper

• The paper is also available at: Siamese graph convolutional network for content based remote sensing image retrieval

• If the work is any help to you, please feel free to cite the author:

@article{chaudhuri2019siamese,
  title={Siamese graph convolutional network for content based remote sensing image retrieval},
  author={Chaudhuri, Ushasi and Banerjee, Biplab and Bhattacharya, Avik},
  journal={Computer Vision and Image Understanding},
  volume={184},
  pages={22--30},
  year={2019},
  publisher={Elsevier}
}