Robust supervised discrete hashing

In this paper, we proposed a more robust supervised hashing framework based on the Cauchy loss function and Supervised Discrete Hashing (SDH) called Robust (RSDH), which can learn subspace consisted of binary codes. The is used to measure error between label matrix product decomposed matrices. RSDH not only reduce outliers noise codes, but also achieve satisfactory retrieval effect. Image experiments demonstrate that performs better than other methods.