Advertisement Detection, Segmentation, and Classification for Newspaper Images and Website Snapshots

Advertisement plays an important role in the human society. Advertisement studies are related to many important issues in economics, social science, and marketing. In this paper, we propose a system to detect, segment, and classify advertisements from newspaper images and website snapshots, in order to facilitate advertisement studies. First, we detect advertisement candidates based on a connected components method. We then design rule-based filters and learning-based filters to remove non-advertisement candidates. From the remained advertisement candidates, we extract visual features and construct classifiers to classify them into predefined advertisement categories. Based on the advertisement categories published over years, we uncover several interesting statistics derived from newspaper front pages and website snapshots.