Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 97 === The digital information over Internet have been growing rapidly recently. Digital multimedia can easily be downloaded or manipulated and tampered intentionally. So the intellectual property protection becomes more important. Most multimedia is stored in compressed bit stream format to save the storage space or transmission time. This thesis proposes two novel techniques for data hiding, data hiding in halftoning images and data hiding in grayscale images, respectively. Notably, the proposed data hiding in grayscale images can also achieve compression benefit.
First, this thesis presents a reasonable computational complexity data hiding algorithm to embed secret pattern into two or more halftone images. The former approaches in the literature do not consider noise strength according to the local variance value. The proposed Adaptive Noise-Balanced Error Diffusion (ANBEDF) method employs the Quality-Noise Look Up Table (QNLUT) and the optimized multipliers to control the adaptive noise strength according to the local variance value. The experimental results show that higher decoding rate is available under the same image quality performance than former approaches in the literature.
Moreover, this thesis proposes another data hiding which can embed flexible amount of data in the bit map of an Ordered Dither Block Trumcation Coding(ODBTC)image, where the ordered dithering is used to dither the quantized BTC image to avoid the annoying false contour and blocking effect inherently existed in BTC image. Experimental results demonstrate that an objective good image quality image with flexible capacity and reasonable complexity is obtained. Moveover, the correct decoding rate of 100% is maintained, and the original host ODBTC image can also be reconstructed in the decoder when needed.
|