A Content-Based Image Retrieval Scheme for Encrypted Domain Using Feature Fusion Deep Supervised Hash
Qiuyu Zhang, Zhen Wang, Xuewen Hu, Ruihong Chen · 2023
To improve the semantic representation of image features and to achieve secure and efficient image retrieval in cloud computing, we proposed a content-based image retrieval scheme for encrypted domains using supervised deep hash function (FFDSH) feature fusion. First, CNNs and VGG16 are used to extract deep features from the image, respectively, and the feature fusion technique is used to merge the two features. The final step is the construction of the deep hash sequence of the fusion features, which is used as the FFDSH sequence. During retrieval, retrieval of the bilayer similarity match is performed according to the obtained FFDSH sequence. The experimental results show that the proposed scheme can achieve efficient image recovery and has apparent benefits in recovery performance compared to existing image recovery schemes.