Domain adaptive video annotation has received significant attention, due to the large increase in digital data. Video Annotation encounters many difficulties, such as insufficiency of training data, curse of dimensionality and the problem of semantic gap. As the manual annotation takes more time and is labor intensive, therefore automatic annotation is highly desirable. In this book, we have proposed a framework for effective video annotation [AVA-VC] which is based on visual content. The Sailency feature extraction technique is used initially, which is followed by shot detection and two level keyframe extraction technique. The proposed feature extraction technique and use of COREL5 image database improves the result of of video annotation. Generation of the weight vector in the training phase and using this newly generated weight vector to find out the annotation, leads in improving the performance. Trecvid dataset is used to test the performance of the proposed algorithm. The proposed AVA-VC outperforms for 38 and 36 concepts on MAP, when it is compared with well known algorithms OMG- SSL and MMT-MGO.