| Summary: | 碩士 === 朝陽科技大學 === 資訊管理系碩士班 === 91 === With the rapid advancement of computer networks and multimedia technology, people can exchange and transmit a good many images through computer networks. However, the secure mechanism of an ordinary computer network, especially a wireless network, is insufficient so that the data transmitted on it would be easily intercepted. Thus, preventing the secret images transmitted on a computer network from being hacked by unauthorized users has become an interesting issue for many researchers.
In a halftone binary image, the patterns with a greater number of black pixels are used to display the dark regions in an image, and those with fewer black pixels are used to delineate the bright regions. Consider a halftone binary image I segmented into small image blocks. When one changes the image blocks to become other image I' where he exchanges the places of the pixels in each identical image block, I' and I still look very much alike. Based on the property, this thesis proposes a halftone-based binary image hiding (HBBIH) to process the halftone binary image embedding.
The proposed HBBIH method can provide a high embedding rate. In this embedding method, even though the error rate between the extracted secret image and the original secret image is great, the extracted secret image still can clearly depict the content of the original secret image. From the experimental results we observe that the error rate is not suitable for describing the unlikeness between two halftone binary images. Besides, if the stego image is damaged while being transmitted on the network, only the secret image blocks whose indices are damaged would be destroyed. Therefore, this method has a robustness capacity in the presence of noise on stego image.
When designing a successful image hiding system, some basic requirements such as imperceptibility, statistically undetectable, security, capacity, and robustness must be taken into consideration. In recent years, many techniques and methods about image hiding have been continually introduced. But, most of these techniques only focus on the former four requirements and neglect the requirement of robustness. Therefore, a robust image hiding method that also contains the feature of robustness is proposed in this thesis.
Generally speaking, when the amount of information that is embedded into a cover image grows, imperceptibility of information relatively becomes worse, and the exposure risk of embedding act becomes higher. In order to increase the capacity of information and also have imperceptibility of it, the information imperceptibility demanded in image hiding technique is not as strict as that demanded in digital watermarks technique. However, due to the limit of bandwidth for information transmission on network, users generally hope that loss data compression is processed in advance onto cover image to reduce the data before data transmission. Furthermore, when data is transmitted via Internet, it may be changed because of the influence from environment. As a result, an excellent image hiding technique should at least have certain degree of robustness against the damage of data that is not from intentionally attacks by others.
In order to increase image hiding capacity, VQ compression technique is used to do lossless compression toward secret images first, and the compressed data is grouped as belonging to several classes. Also, the method of discrete wavelet transformation — DWT is used to transform a cover image represented by space domain into a frequency domain image presented by frequency domain. Then, the data of some class in a secret image is embedded into the lower frequency sub-band that is not easily changed in a frequency domain image, while the part of other classes is embedded into middle frequency sub-band. Besides, when the compressed data of certain block B in a secret image is damaged, the system will recover them by using the blocks adjacent to B, which are hided in the lower frequency sub-band.
|
|---|
