Copyright infringement has emerged as a significant issue in the growth of the e-book/comics market on account of illegal copying and distribution. It is therefore important to develop automated comic book identification techniques to prevent such problems. Fingerprinting methods have been typically used for multimedia identification; however previous fingerprinting methods are unsuitable for the identification of comic books, which may include several types of distortions, including geometric changes. In this paper, a new comic fingerprinting method is proposed based on a comparison of the average pixel intensity of sub-images, which are called pairwise patterns. In particular, for robust identification against such geometric distortions, circular patterns are newly proposed and evaluated after constructing a comic fingerprint database. The intra- and inter-distances of the features in a fingerprint are calculated to demonstrate its pairwise independence and robustness against various distortions. The result shows that the proposed fingerprinting method is more robust against various distortions, especially for rotational distortion, than those in previous methods. Moreover, owing to its low complexity, the method has potential advantages for commercial applications in real-time.1
Bibliographical notePublisher Copyright:
© 2015 Elsevier B.V.
All Science Journal Classification (ASJC) codes
- Signal Processing
- Computer Vision and Pattern Recognition
- Electrical and Electronic Engineering