Feature Analysis and Classification for Filtering Junk Information in Animation
Ming Nan Zhao, Shilin Wang, Shenghong Li, Xiang Li, Zhi Xue · 2007
Digital animation is a widely used digital media on Internet to convey information. However, many animations nowadays are usually advertisements and contain only junk information. In order to detect and filter such information, a feature extraction, analysis and classification method for animation content understanding is proposed. A feature set composed of the traditional image/video features and other specific features for animation is extracted. Then a feature analysis method based on Mutual Information (MI) is performed to select the feature combination with high discriminative power. Finally, SVM with RBF kernel is used as the classifier and an average error of 8.28% is achieved by the optimum feature set.