Hiding Approach for Efficient Image Indexing

In this letter, we propose a data hiding approach to embed the indexing keys inside the JPEG compressed images for their retrieval, inspired from the spirit of digital watermarking. At the stage of database population, the existing approach is to compress all images by standard JPEG in order to save their storage space, but ignored the compression of their indexing keys under the notion that the size of each indexing key is negligible. When a large image database is established, however, this part of storage space becomes non-trivial. By hiding the indexing keys inside the JPEG compressed codes, significant advantages are gained that the indexing keys can be compressed naturally by JPEG without any additional cost. Our experiments on hiding an indexing key of 256 bytes illustrate that: (a) no noticeable distortion is introduced in comparison with non-watermarked images; and (b) their storage space is reduced into the region of 7%~27% of their original size.