Improved CNN-Based Hashing for Encrypted Image Retrieval

As more and image data are stored in the encrypted form cloud computing environment, it has become an urgent problem that how to efficiently retrieve images on encryption domain. Recently, Convolutional Neural Network (CNN) features have achieved promising performance field of retrieval, but high dimension CNN will cause low retrieval efficiency. Also, is not suitable directly apply them for To solve above issues, this paper proposes improved CNN-based hashing method retrieval. First, size increased inputted into improve representation ability. Then, a lightweight module introduced replace part modules reduce parameters computational cost. Finally, hash layer added generate compact binary code. In process, code used which greatly improves The experimental results show scheme allows effective efficient images.