With the proliferation of camera phones, new information retrieval applications will emerge. The image of a scene captured by a camera phone can be a query to a remote server to identify the scene and return relevant information. But unconstrained scene identification is an open problem. In this paper, we propose a discriminative measure to rank image patterns sampled from target scene classes. Support vector classifiers are then trained using top discriminative patterns for scene identification using voting. We demonstrate our generic approach on two scene databases (ZuBuD and STOIC) with promising results.