Classification of texture based on Bag-of-Visual-Words through complex networks

Abstract Over the last years complex data (e.g. images) have been growing in a very fast pace. This demands the ability to describe and to categorize them. To solve this problem it is essential to develop efficient and effective vision-based expert techniques. Hence, the cornerstone of our work is to propose a new methodology, called BoVW-CN, that combines Bag-of-Visual-Words and complex networks for describing keypoints detected in a given image. Our insight is that describing just the relevant points of an image we can achieve a more cost-effective and better image description. The obtained results testify that BoVW-CN, applied to public image datasets, outperforms the widely used state-of-the-art methods. We not only obtained good accuracies (e.g. 78.18%), but also performed analyses to find the best trade-off between computational cost and accuracy. Besides, to the best of our knowledge, our work is the first one to propose such integration of Bag-of-Visual-Words and complex networks through a texture-based focus.