Color Image Digital Watermarking Based On C-IQA And Visual Perception

Information technology is developing and growing rapidly, which makes it easy for digital information to spread. It tremendously boosts the efficiency of our studying and enriches our daily life. Among various digital information, the digital images carry much information, and they are the most widely spread. However, due to its distortion-free copying and wide-spread dissemination, the copyright protection, ownership declaration and file integrity of digital image is under great threat. How to protect the copyright of digital information, especially the digital images, attracts more and more attention. The most widely used method to protect the image copyright is digital watermark. However most of the current watermark algorithms are performed in gray-level images; the methods for color images are not quite mature. So, this thesis is mainly focused on the color image watermark.The existing color image watermark techniques are mostly the extended version of the techniques for gray images, and seldom focus on the features of the original color images, leading to the fact that the watermark schemes don’t combine the quality and visual perception of the original image well with the design of watermark schemes. Considering the current situation discussed above, this thesis analyses the features of the original color carrier images and focuses on the image quality assessment and the perception of human visual system, and proposes an adaptive digital watermark algorithm for color image based on the color image quality assessment and the perception of human visual system. In order to take the overall characteristics of the original carrier image into consideration, this thesis analyses its color and gray features respectively. Color Root Mean Enhancement (CRME) method is a kind of color quality assessment, which is based on the contrast of the color images and analyzes the color variance and structural differences to give the color images quality assessment. The higher the calculated value is, the better the perceived image quality is, and vice versa. Watson model calculated the threshold of the changes of DCT coefficients in gray-level that human eyes could perceive. The higher the threshold is, the lower the visual sensitivity is, and vice versa. Hence, the algorithm takes advantage of CRME method to measure the perceived quality, and at the same time takes advantage of the Watson visual perception model to evaluate the gray-scale sensitivity.In order to avoid the uneven distribution of image quality, the proposed algorithm divides the image into non-overlapped blocks. After the carrier image is transformed from RGB color space to HSI color space, the algorithm embeds the watermark in the block-wise DCT coefficients of the H (hue) component; the embedding strength is adaptively set:if the perceived quality is high and the perceptual threshold is large, the embedding strength is set big, and vice versa.In order to prove the feasibility of the proposed algorithm, we’ve conducted many experiments. We’ve embedded and extracted the watermark under the attacks of different intensity of noise, filtering and geometric cropping. The experimental results show that the color image watermarking algorithm proposed by this thesis can keep the quality of the images maintain a high level, and be able to stand the attacks of different intensity of noise, filtering and geometric cropping, which means that the algorithm is robust. The algorithm can be applied to color image copyright protection in related fields.