Image Meta-feature Extractor

Library to extract meta-features from images.

Instalation

git clone https://github.com/gabrieljaguiar/image-meta-feature-extractor.git

Instructions

The input folder is the folder of every image the user wants to extract the meta-features. The images can be in any format that opencv can open. The output file must be a .csv file in which the features are going to be write. There is a folder with example files, please check them.

python run.py ./example/input_folder/ ./example/output/output.csv/

Meta-Features

In this library, 97 meta-features are extracted.

• Statistical (3)
• Colour-based (36)
• Histogram (21)
• Border (16)
• Image Quality (2)
• Texture (19)

Requirements

The following Python packages are required:

• numpy
• pandas
• opencv2
• scikit-image
• imutils

Also, use Python 3.6+!

Citation

All of these features are presented or referenced in Aguiar et al (2019) [1]. Also, if you use this extractor, please cite us:

@article{aguiar2019meta,
  title={A meta-learning approach for selecting image segmentation algorithm},
  author={Aguiar, Gabriel Jonas and Mantovani, Rafael Gomes and Mastelini, Saulo M and de Carvalho, Andre CPFL and Campos, Gabriel FC and Junior, Sylvio Barbon},
  journal={Pattern Recognition Letters},
  volume={128},
  pages={480--487},
  year={2019},
  publisher={Elsevier}
}