Sparse graph based self-supervised hashing for scalable image retrieval

•A Sparse Graph based Self-supervised Hashing (SGSH) method is proposed to solve the performance degradation problem in traditional graph based hashing method;•SGSH can not only circumvent the requirement of the pre-defined category number, but also alleviate the problem of large memory demand;•Self-supervised reconstruction constraint is proposed to further preserve semantic information for binary codes;•Extensive experiments on four popular datasets demonstrate the superiority of the proposed method.