Robust Digital Watermarking Method Based On Image Defocus Attribute

With the help of the integrated media construction,the quality and efficiency of image dissemination have been greatly improved,which improve people's requirements.With the development of photography technology,people can take exquisite image easily.They place the objects on different visual levels through background blur technology,which can not only highlight the thoughts they want to express,but also enhance the aesthetic effect of the picture.Nowadays,these images with beautiful defocus attribute have been widely disseminated.However,it will be an infringement to reprint these images without permission by the copyright owner or even use it as commercial purposes.Copyright awareness is very important today and should be properly resolved.The digital watermarking technology which originated in the 20 th century,makes full effect in information security issues such as copyright protection.In this digital watermarking technology,the method of incorporating the Just Noticeable Distortion(JND)calculation has been deeply studied.But there is still a lack of research on how to reach a high level in visual effect and robustness.It is noticeable that there is no visual model that incorporates the defocus attribute.The defocus attribute is similar to the visual saliency of classic images.Thus,it can also be used in the JND calculation of classic images.On the one hand,the human eyes pay attention to certain areas easily while it will be emphasized by defocus attribute,so the slight distortion in these areas may reduce the overall viewing effect of the image,while there will be overestimation using the ordinary JND.On the other hand,the calculated JND of the watermarked image after some attacks is likely to be quite different from the JND of the original image.Thus,the above effects are different from the influence of saliency on JND.Based on this,we propose the robust digital watermarking method considering the defocus attribute,which mainly includes two tasks:1.Digital watermarking algorithm based on defocus attribute JND modelThe JND threshold will be estimated to be high or low due to the phenomenon that the human eyes pay different attention to different focus and defocus areas in the image.Thus,we analyze the influence of defocus on block energy and the surrounding image blocks to the central image.The visual influence of the block is finally calculated by different image types,and then a new contrast masking(CM)calculation method is constructed based on spatial frequency masking,thereby forming a new JND model.Then we apply it to digital watermarking technology using spread transform dither modulation(STDM).The experimental results in defocus images and classic images prove that our proposed watermarking method based on the defocus analysis JND model is more robust than some common JND models and watermarking methods while maintaining the same visual quality.2.Robust digital watermarking method based on saliency modulation calculated by defocus attributeThere are two possible improvements in the first watermarking method: the division of image block types may lead to inconsistent judgments on the watermarked image with attacks;this watermarking method is not robust to JPEG compression attacks as robust to other attacks such as noise.In order to further solve these two problems and improve the general applicability of our proposed method,the defocus attribute is used as the starting point of the study.In detail,we combine the sharpness characteristic that has a positive stimulating effect on vision and the texture complexity that has the reverse masking effect in the calculation of image saliency.At the same time,the new CM value is calculated.In addition,the impact of color complexity is considered.In this way,a new JND model is formed.This JND model is also applied to the STDM watermarking method.The experimental results prove that the robustness of this method in noise,compression,shearing,rotation,filtering,and scale attacks is better than some existing watermarking methods while maintaining the same visual quality.