Hashing is a key technique to achieve fast nearest neighbor search in high-dimensional

massive datasets.Among various methods

Iterative Quantization (ITQ) and Isotropic Hash (IsoHash) are probably the most popular ones due to their high retrieval accuracy.However

as the constraints imposed on the rotation matrix are too weak

the optimization problem in ITQ is severely under-deterministic and therefore easy to cause over-fitting.In IsoHash

the isotropic projection matrix is updated in a manner that is completely independent of the binary hash codes

and thereby the quality of the produced hash codes may be depressed.To address these issues

this paper proposes an isotropic iterative quantization hashing method

which extends the formulation of ITQ by incorporating properly the isotropic prior adopted in IsoHash.In our method

the hash code matrix and rotation matrix are updated alternately in an iterative fashion.Experiments are conducted on three benchmark datasets

CIFAR-10

22K LabelMel and ANN_GIST_1M.The results show that the proposed method performs better than the competing methods in terms of precision

recall and mAP.