Compact Structure Hashing via Sparse and Similarity Preserving Embedding

Over the past few years, fast approximate nearest neighbor (ANN) search is desirable or essential, e.g., in huge databases, and therefore many hashing-based ANN techniques have been presented to return the nearest neighbors of a given query from huge databases. Hashing-based ANN techniques have become popular due to its low memory cost and good computational complexity. Recently, most of hashing methods have realized the importance of the relationship of the data and exploited the different structure of data to improve retrieval performance. However, a limitation of the aforementioned methods is that the sparse reconstructive relationship of the data is neglected. In this case, few methods can find the discriminating power and own the local properties of the data for learning compact and effective hash codes. To take this crucial issue into account, this paper proposes a method named special structure-based hashing (SSBH). SSBH can preserve the underlying geometric information among the data, and exploit the prior information that there exists sparse reconstructive relationship of the data, for learning compact and effective hash codes. Upon extensive experimental results, SSBH is demonstrated to be more robust and more effective than state-of-the-art hashing methods.