| Summary: | 碩士 === 國立金門大學 === 電子工程學系碩士班 === 107 === This thesis uses data hiding as the research topic, which is divided into steganography and digitial watermarking. The former research is that the amount of secret information will affect the image quality without considering the transmission process. In the study of the damage, the latter study is to verify the identity of the image when the image message is destroyed. In this paper, two different ways of steganography are proposed by steganography.
In this paper, the first steganography method uses a block function in the color image hiding method, using a private key as a random seed.The 2×2×2 block structure with random 0 to 7 values is combined into a matrix model of 8×8×8 block functions, and copied into a matrix model of 256×256×256 large block functions, and this large block The matrix model of the function is quite complex and highly secure and difficult to crack.The secret image embedding stage is to put the original image pixel value (RGB) into the reference coordinate of the matrix model of the large block function, and modify the reference coordinate of the matrix model of the large block function to which the secret image is located in the rule of minimum distortion. position.In the secret image extraction stage, the modified pixel of the hidden image is directly extracted from the embedded secret image by using the matrix model reference coordinates of the large square function. The experimental results show that the proposed method preserves the binary color image after the jitter processing in the original image and maintains good image quality, and the average PSNR value is 51.14 dBs.
In this paper, the second steganography method uses the selective block method to improve the steganography method for the Absolute moment block truncation coding for the 7,4 Hamming code. The absolute momentum block truncation code is a kind. An effective image compression method. When the 7,4 Hamming code hides the secret message in the absolute momentum block truncation code, the PSNR of the hidden image is significantly reduced. This problem stems from the inherent characteristics of the AMBTC method. In the compressed code, a compression code of 1 indicates a high average value, and a compression code of 0 indicates a low average value. When 1 is switched to 0, the high average value in the decompressed image is replaced by a low average value. The high average and the low average vary between blocks and blocks. Some replacements can result in severe grayscale changes in the decompressed image, greatly reducing the PSNR value. In this paper, a new method for improving the hidden image of the 7,4 Hamming code in the absolute momentum block truncation code image is proposed.
The difference between the high and low average values in all coding blocks was calculated and the statistical distribution of the differences was analyzed. The difference value is truncated to the appropriate threshold block according to the size of the secret image. The secret image is hidden in the low difference block of the AMBTC image. In order to obtain a better hidden capacity, the secret image is pre-quantized into a binary dithered image. In the experiment, the original image is a 512x512 hidden image with a hidden size of 256x256. The average PSNR of the hidden image is 33.1703dBs and the average PSNR value of the AMBTC image is 33.1805dBs, which is about 0.01dBs.
|
|---|
