| Summary: | 博士 === 國立中正大學 === 資訊工程研究所 === 99 === Data hiding can be used to assist secret message communication on the Internet. The secret message is first encrypted with a traditional cryptographic method and then embedded in a cover or host media through some data hiding algorithm. The stego-media (host media with embedded data) would then be sent to a receiver on the Internet. The stego-media must be perceptually similar to the original host media. This can greatly reduce the risk of secret messages being inspected by network hackers.
The study of large payload data hiding techniques will evaluate data hiding in different image hiding domains, including simple Least Significant Bit (LSB) substitution, Vector Quantization (VQ) compression and reversible data hiding domains. The focus will be on the ability to choose better locations in a host image for data hiding. The stego-image will have better quality (in terms of PSNR values) under the same hiding payload and the total hiding payload can be larger than for previous work.
In the LSB substitution based data hiding, we proposed a technique called “Variable Rate Data Hiding by Simple LSB Substitution and Quadtree Segmentation” that will embed more data in the complicated area and less data in the smooth area. A quadtree segmentation technique was used to separate the host image into complicated and smooth areas. The stego-image’s quality was enhanced, especially in the smooth area.
Images are often transmitted on the Internet in compressed format to decrease the consumption of communication bandwidth. Therefore, it would be desirable to hide data in these reduced (compressed) bitstreams. Data can be embedded in an image’s VQ compression codes by modifying the VQ indices. Each VQ index can thus embed 1 bit of data. We proposed a technique called “Data Hiding on Two-stage VQ Compression Codes” that theoretically doubles the data hiding payload under the VQ compression domain. Also, in the VQ domain, we proposed a second method called “A Best-pair-first Capacity-distortion Control for Data Hiding on VQ Compression Domain” that will always pick the best VQ location for data hiding. A higher stego-image’s quality can be obtained under any hiding capacity compared to the traditional data hiding method on the VQ compressed domain.
In the final method, we investigated the reversible data hiding technique, which allows the stego-image to be recovered back to its original host image after the embedded data is extracted. This is a desirable feature for data hiding applications in areas like the legal documents, military images, and medical images, where the host image’s quality cannot be compromised. We proposed a technique called “Using Quad Smoothness to Efficiently Control Capacity-Distortion of Reversible Data Hiding” by enhancing the difference expansion (DE) based reversible data hiding scheme to efficiently choose better locations (quads) in which to embed data. Experimental results showed higher stego-image quality and increased data hiding payload in comparison with other DE-based reversible data hiding methods.
Our future work will continue to focus on enhancing a stego-image’s quality and/or increasing the data hiding payload. The ability to choose better locations for data hiding will continue to be our major concern. For traditional data hiding in the spatial domain, we will try to enhance the hiding security on LSB-based data hiding. For the VQ compression domain, we could try to combine the ability to choose better locations for data hiding and to increase the data hiding payload. Finally, the reversible data hiding method may be used in applications for medical images.
|
|---|
