Publication Date: 2024/02/12
Abstract: With the proliferation of digital images, there has been considerable research into Contentbased Image Retrieval (CBIR) techniques. Typically, CBIR services demand substantial computational and storage resources, making it advantageous to outsource these services to cloud servers equipped with abundant resources. However, the challenge arises in ensuring privacy, given the inherent lack of complete trust in cloud servers. In here, we introduce a delegated CBIR approach adapted from a book BOEW model. Our method involves encrypting images through hue value replacement, block and intra-block pixel permutation. Subsequently, cloud server calculates regional histograms from these protected image blocks, binds and employs the resulting cluster centers as encrypted viewable words. It is this approach allows us to construct a Bagof-Encrypted-Words (BOEW) model, representing each hue as a feature vector—specifically, a generalized histogram of the encrypted viewable words. To measure the resemblance between images, we utilize the Manhattan space between feature vectors on the cloud server. Experimental output and a security analysis of our proffered scheme illustrate its search precision and security.
Keywords: No Keywords Available
DOI: https://doi.org/10.5281/zenodo.10650100
PDF: https://ijirst.demo4.arinfotech.co/assets/upload/files/IJISRT24JAN415.pdf
REFERENCES