FIT Ph.D student published an academic article in a top journal within the field of multimedia that has attracted the attention of CCF TCMT

FIT Ph.D student published an academic article in a top journal within the field of multimedia that has attracted the attention of CCF TCMT

Recently, Wenguang He, a Ph.D. student of the Faculty of Information Technology, Macau University of Science and Technology, published an article entitled “High-fidelity Reversible Image Watermarking Based on Effective Prediction Error-Pairs Modification” in the top international academic journal IEEE Transactions on Multimedia (JCR Ranking:5/108) under the supervision of Prof. Zhanchuan Cai. The latest impact factor of the journal is 6.051. After the article was officially published in January 2021, it attracted the attention of CCF TCMT, and the technical committee believed that the article has novel ideas, forward-looking methods, and detailed experiments, and hoped that it can be introduced through the official channels of the technical committee to benefit more researchers. The link to the introduction to academic articles is https://mp.weixin.qq.com/s/pF3Ey5tuOdRjN0q7uNSEFQ.

As an important approach in the field of information security, data hiding generally introduces permanent distortion into the cover medium. To settle this issue, an important branch called reversible data hiding (RDH) proposes to exactly restore the cover medium after data extraction. The property of reversibility allows it to deal with the secret transmission or content authentication for sensitive image processing. Prediction error expansion (PEE) is the mainstream technique of exploiting the redundancy of the cover medium to realize RDH. To achieve adaptive embedding in pixel value ordering (PVO) based 2D PEE, a new RDH scheme is designed based on the observation that the distribution of error-pairs has a fundamental impact on the performance of 2D PEE. Firstly, adaptive prediction error generation is designed to optimize the distribution of error-pairs. Existing research shows that in PVO the prediction target selection based on pixel value is not conducive to the generation of expandable error and the relative location should be introduced. By introducing the relative location of the largest/smallest three pixels within block into prediction, the distribution of error-pairs is optimized. After that, flexible spatial location is adopted to achieve further optimization. Specifically, since only the largest/smallest two pixels within block are modified during data embedding, they can be replaced by the third largest/smallest pixel to obtain an invariant block. Next, every candidate mode of defining spatial location is applied to obtain the location sequence and the inverse number. Finally, the mode with the largest inverse number wins. Adaptive prediction-error has also been achieved. The above work which mainly focuses on expandable error-pair is based on the assumption that two pairing errors are both modifiable. Taking into account the high similarity between pairing errors, if one of them is shiftable the other one is also shiftable with high probability. To avoid shifting two pairing errors, a new mapping that is more suitable for non-smooth blocks is derived from the existing one. Compared with existing adaptive embedding schemes, not only adaptive prediction error generation but also adaptive prediction error modification has been realized in the proposed one. The idea of fully-adaptive embedding has definite significance for the further study of PEE.

The study was supported by the Science and Technology Development Fund of Macao SAR under Grant No. 0012/2018/A1 and Grant No. 0069/2018/A2. The link to download the article is https://ieeexplore.ieee.org/document/9042201.

Captured from the introduction to academic articles of the technical committee