Abstract: With the raising in the amount of images in image matching and retrieval in social Media , Image annotation has egressed as an significant research issues due to its practical application .Many social image search engines are based on keyword/tag matching . In this paper , the proposed method combines the prediction models for different tags into a matrix, and casts tag ranking into a matrix recovery problem. It acquaints the matrix trace norm to explicitly control the model complexity, so that a reliable prediction model can be acquired for tag ranking even when the tag space is big.

Keywords: Automatic Image Annotation; Tag Ranking; User Tags; Image Retrieval.