The popular smart phone and other mobile devices accelerate the use of cloud computing. Cloud can not only bring to the user resources for computing and application but also offer new opportunity for solving difficult problems. A novel method aimed to bridge semantic gap is proposed in this paper. Users can access the wanted query results by labeling multimedia data in cloud with people's collective intelligence and similarity search by virtue of high performance of cloud. The multimedia data can be labeled by people's endeavor with highly accurate semantic meaning. The huge amount of near duplicated data can be clustered by similarity search efficiently.