Research On Quality Evaluation Of Lithographic Watermarked Image Based On Image Dissimilarity

The rapid development of commodity economy and the gradualimprovement of the property rights have encouraged people to seekefficient anti-counterfeiting technology. It would be used to combatincreasingly rampant counterfeit behavior in the modern. Because ofthese advantages of difficult to copy, easy to identify, bottom cost andpollution-free, the digital watermarking technology has became animportant research and development direction in the field of packagingand printing anti-counterfeiting as a new means of security. The imagequality assessment algorithm is a key link in digital watermark anti-counterfeiting technology, which plays an important role to control thequality of watermarked image and print. Hence, after researching theexisting algorithms, the paper put forward an image quality assessmentmethod based on the printing ink density and chromaticity from theperspective of otherness. And aimed at a holographic watermarkingembedding algorithm based on DWT, the evaluation standard wasdetermined to match the quality inspection results in printing industry.Firstly, image color feature, texture feature and shape feature and thecharacteristics of spatial relations are analyzed, which could be used torepresent the difference between images. Based on this, the paper raisedthe concept of image otherness measure. Then singular valuedecomposition (SVD) is chosen as the method of image principalcomponent analysis. Through studying the relationship of between SVDand image structure and energy information by the matrix norm theory, itcould be found that the deviation degree of feature vector (singular values)could describe different features.Secondly, by using the block extraction method based on thecharacteristic of image spatial relationships, a watermarked image qualityevaluation method based on Singular Value was put forward. It choosesthe correlation of SV diagonal matrix between the host image anddistorted image as a measure of image dissimilarity. And matrix F normof blocks to the sum of the F norm ratio is the weight. The algorithmdefines the distortion degree of low frequency and high frequency for SVQIL(Singular Value Quality Index) and SVQIH, respectively.Meanwhile, according to aberration and density difference measured byprint apparent quality detection, threshold of the RGB images SVQI withgood visual perception is set for SVQIL∈(0,15)，SVQIH> SVQIL.Finally, four degradation simulation experiments were the proposedalgorithm. It could be found that SVQILand SVQIHcan both reflect thedegradation of the image. The evaluation results not only accord withhuman visual perception, but also better than the PSNR and UIQI. Andthen through a series of print/lithographic-scanning experiment, it wasstudied that the watermark embedding and printing parameters exercisethe influence on the image quality. The experimental results show that:with the embedding strength increases, the concealment of the watermarkis weakened, and robustness is enhanced; When the watermark isembedded into R channel, the concealment of the watermarking is better;the robustness of watermark is better as the watermark embedded into themost color component channel of the RGB image; Paper surface featuresnot only have obvious effect on printing quality, also has a significantinfluence of the recovered watermarking; When300g/m2coated paper isused as the substrate, the quality of the watermark image and recoveredwatermarking is better than200g/m2coated paper; Scanning resolutionto has a little influence on the quality of the recovered watermark, butcompared to300dpi scan resolution, the600dpi scanning quality oflithographic watermarked image is better.In this paper, the beneficial exploration has been carried on the qualityevaluation of lithographic watermarked image based on imagedissimilarity. The results obtained would provide support and help for thedevelopment of digital watermark anti-counterfeiting technology.