期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2014
卷号：3
期号：6
页码：2224-2228
出版社：Shri Pannalal Research Institute of Technolgy
摘要：Automatically assigning relevant keywords for an image provides easier retrieval of images and understanding of large collection of image data. In this paper, a new image annotation algorithm was introduced which is used for image retrieval process. The proposed system is based on the similarity measure between images which is referred as compactness. The compactness based matching model shows how close the test image features lie to the training image cluster centers. The usage of low level features maintains low running time and gives more accurate detail information. At annotation time similar images are retrieved from the database using similarity measure and the labels from these images are used to form the annotation. This method makes use of different label transfer techniques to transfer the labels from the retrieved images. This annotation process requires only simple training algorithm and can efficiently work with different low level image features.
