AUTOMATIC VIDEO ANNOTATION BY CLASSIFICATION
Manjary P. Gangan, R. Karthi · International journal of advanced research in computer science and electronics engineering · 2012
Automatic video annotation is a technique to provide semantic video retrieval. The proposed method of automatic video annotation consists of two main steps i.e.; feature extraction and algorithm for annotation. The features are extracted from the images in the database and these feature vectors are then provided as a training set to the classifier algorithm. The classifier algorithm is a combination of graph based algorithm and K nearest Neighbor algorithm. In graph based algorithm, the computed feature vectors of each image in the database are taken as nodes of the graph and the neighbourhood information of each node is used for classification. In KNN, the classification is using majority vote among the K objects. When a query video is provided by the user, the key frame is extracted and, pre-processing and feature extraction is performed on this key frame of the video. The extracted feature is then provided to the trained algorithm for annotation. The proposed system has a precision rate of 81.14%. The system also compares the results of different combinations of features and descriptors. An automatic annotation system can be used for effective search and retrieval of news videos, event detection, video summarization and highlight generation, video content analysis, etc.