Research On Multilevel Halftone Image Generation And Copyright Protection Of Ceramic Tile

The precision requirement of high-end ceramic tile to pattern is very high.Generally,the pattern is printed on the ceramic tile blank by industrial 3D printer and then fired to complete.The industrial 3D printer has 6 ～ 12 channels,and each color nozzle has 4 quantization levels.The ceramic tile design drawing in the standard image format is divided into 6 ～ 12 channels,corresponding to the printer to obtain the color separation image(256 quantization levels).Then the color separation image is converted into the multilevel halftone image(4 quantization levels)required by the color nozzle.This process is called the halftone process.According to surveys,the color separation and multilevel halftone image generation technology,and its system is grasped by the foreign companies.Therefore,this kind of key technology brings uncontrollable factors to the system upgrading multi-functional integration,which limiting of the development of relevant domestic technologies.In addition,the theft of ceramic tile halftone image will bring immeasurable economic losses to ceramic tile factories,so it is imperative to protect the copyright of ceramic tile halftone images.There is little research on the above problems in China.I participated in the research topic of ceramic tile traceability based on information hiding by my tutor,and proposed the generation algorithm of ceramic tile multilevel halftone image based on improved U-Net and error diffusion,the classification algorithm of ceramic tile texture complexity based on the reverse model of color separation image and the watermark algorithm of multilevel halftone image based on ceramic tile texture complexity.The experimental results have been effectively verified in the actual proofing test of Guangdong Dongpeng ceramic tile factory(Qingyuan base).The main research work of this thesis is as follows:(1)According to the requirements of high fidelity and high precision of ceramic tile design drawing printing,a multilevel halftone image generation algorithm of ceramic tile based on improved U-Net and error diffusion is proposed.Firstly,in order to solve the problem of non convergence in model training caused by the imbalance of tile image data,U-Net network is selected to pre train the self built landscape image data set,and the input and output size,loss function and activation function of the network are improved to make it suitable for the prediction of tile multilevel halftone image.Secondly,the trained model is transferred to the tile image data set,and the tile halftone image is segmented and predicted to obtain the block set of continuous tone transition image.Thirdly,each transition image is pieced together to form a complete continuous tone transition image.Finally,the improved error diffusion method is used to generate the multilevel halftone image of ceramic tile.The algorithm can convert the color separation image of ceramic tile into multilevel halftone image,and ensure that the ceramic tile pattern generated by printing,firing and other processes is basically consistent with the original design pattern.Its actual proofing quality is recognized by the designers of ceramic tile factory,which is feasible.(2)The classification accuracy of texture complexity of multilevel halftone image affects the quality of image after information hiding.The data is scattered and the numerical range of ceramic tile halftone image is small,it is difficult to describe the texture characteristics of ceramic tile.Therefore,a ceramic tile texture complexity classification algorithm based on the reverse model of color separation image is proposed,by establishing the mechanism of texture complexity classification of color separation image corresponding to that of halftone image.Firstly,the improved U-Net network is trained through the landscape image data set,and the input and output size,activation function and convolution kernel of the network are improved to make it suitable for the reverse rendering of tile color separation image.Secondly,the training model is transferred to the tile halftone image to realize the conversion from tile halftone image to tile color separation image.Thirdly,the texture features of ceramic tile color separation image are extracted,and the gray level co-occurrence matrix and image standard deviation are used as the indicators of texture complexity classification.Finally,four different classification strategies(artificial threshold selection,logistic regression,K-means clustering and binary classification network based on deep learning)are used to binary classify the texture features of the color separation image to obtain the texture complexity classification of the color separation image,that is,the texture complexity classification of the corresponding halftone image.The algorithm can effectively classify the texture complexity of ceramic tiles for subsequent copyright protection.(3)Aiming at the copyright protection of multilevel halftone image of ceramic tile,a multilevel halftone image watermarking algorithm based on ceramic tile texture complexity is proposed.Since the quantization level of tile multilevel halftone image is only 4 bits,the watermark information is embedded into the area with complex tile texture to avoid the decline of tile image quality.Firstly,the tile multilevel halftone image is divided into blocks to form a halftone image block set,and the halftone image block set is rendered through the reverse rendering model to generate a color separation image block set.Secondly,by classifying the texture complexity of each color separation image block,the corresponding halftone image block is selected according to the texture complexity block for watermark embedding,and the sum of the three channels of the block is subjected to modulo-2 operation.The operation result is compared with the embedded information,and the watermark is embedded by flipping the pixels.Finally,all halftone block sets with and without watermark are integrated to form a halftone image embedded with watermark.The algorithm not only effectively protects the copyright of the multilevel halftone image of ceramic tile,but also takes into account the quality of the printed ceramic tile image.