Image Hashing Based on SIFT Features
Pingyuan Li, Xiaoguang Yuan, Suiping Jiang · 2021
Image hashing maps high-dimensional image to low-dimensional binary code which can be used for fast image retrieval. A new image hashing method is proposed in this paper. Firstly, the SIFT features of an image are extracted. Secondly, the SIFT feature's 128 dimensional components are divided into 16 groups and the differential SIFT feature of each group is calculated in a sequential and cyclic way. Thirdly, the weight of each SIFT feature is calculated and is applied to the corresponding differential SIFT feature. Finally, the weighted differential SIFT features are summed to get the image's SIFT feature, and binarization is carried out on each group of the image's SIFT feature in a sequential and cyclic way to obtain the hash code of the image. The proposed method has the advantage that it can quickly generate hash codes for large-scale data sets and is robust to rotation and noise. The experimental results on caltech-256 showed that the proposed method can improve the effect of image retrieval.