| For a long time, high-quality and high-speed plate-setter have been a bottleneck during the development of gravure industry gravure engraving field of key technologies to be solved. Access to high-quality output image is an effective way to improve the level of intaglio plate technology. Laser carving technology has the advantages of higher resolution, focusing precision and digital modulation. Direct laser micro-structure technology has replaced the traditional intaglio plate technology. It has been widely used in laser engraving gravure plate and laser plate processing. Laser engraving carving is a kind of imitation print output device scanning engraving method. Gray scale image halftone algorithm plays a crucial role on laser image output effect beyond of the laser wavelength, pulse frequency, quality of laser beam and processing properties of materials and other factors. Digital image halftone is a continuous tone image such as laser engraving machine, digital printing machine, laser printers and other equipment on the development and value of two in the human visual system to generate a continuous tone image illusion of key technology. In the process of production, it gets more and more extensive application. At present, digital halftone technology is ubiquitous from the home office small desktop inkjet, laser printer, laser engraving machine to large publishing and printing system.In this thesis, the laser engraving the gravure cylinder was studied. The actual proofing applications demand for the technology of laser engraving intaglio. Digital halftone technology as well as the establishment of laser engraving gravure quality of cells and focusing on the detection and sorting system to adapt to the laser engraving gray-scale images using MATLAB7.0.1simulation software, German SCHEPERS laser engraving machine, MZD0745C zoom monocular video microscope intake instrument and SH-DXA gravure proffer and Epson Perfection V300.Firstly, the laser engraving gravure plate-making process output image quality is unstable and plate-making speed by the speed of the image processing time limit, gray-scale images digital halftone algorithm, new the algorithm references and application of the K-means clustering method, the weighted least squares method, the Harr two-dimensional discrete wavelet transform and improved direct binary search algorithm techniques. For the improvement of the quality of the image smoothing area is proposed based on K-means clustering sub-regional digital halftone technology to improve the quality of halftone image, reducing the computational complexity, reduces the computation time required. K-means clustering algorithm on the selection of the initial centers is very sensitive, the algorithm is easy to fall into local optimal solution, so the improved K-means clustering algorithm for image segmentation, the segmented image is more accurate, using the weighted least squares the method of iterative optimization, the iteration speed has been further improved to accelerate the speed of image halftone. Only consider the application of K-means clustering method and the improved method image smoothing area, without considering the image edge information, so the use of the wavelet domain multi-scale information fusion method to get the image edge information, the halftone image quality and speed of plate making a certain improvement was obtained.The results show that the sub-region strategy based on the K-means clustering method of iterative optimization occurs within the region, which the area twice iterative error less iteration then this area stops. Halftoning algorithms constructed with the least squares algorithm based on the model, compared with the increase in the clustering partitions, image smoothing, and the sharpness is increased, especially in the image detail parts. Compared with the least squares model-based algorithm, the mean square error is reduced0.0129to0.2102, the signal-to-noise ratio of the weight is increased from1.4938to2.2072, the peak signal-to-noise ratio is increased from0.53to3.17. The simulation results verify the effectiveness of the algorithm. The application improved K-means clustering gray scale image segmentation by random error term heteroscedasticity, which application of weighted least squares method to build a new energy function, each divided region pixel variance countdown, found by analyzing the pixel mean and variance of the regional as the weight factor, and conversion of the half-tone image. After the calculation of the4iterations of the algorithm, the convergence error drops below0.20, and has a faster convergence rate. Experimental image frequency domain analysis and the results show that with increasing clustering partition, the "bright spots" on the spectrogram gradually tend to be evenly distributed, the proposed algorithm halftone human visual effect is better than the least squares model-based algorithm gray scale values evenly distributed, the image smooth transition. Multiscale information fusion halftone method, mixed error measure function of the establishment of multiscale wavelet coefficients fusion to establish the boundary error measure function and improved K-means clustering method to establish the area error metric function. The applications direct binary search method to minimize the error of the initial image and halftone images, the least squares model-based algorithm in iteration15when convergence error is reduced to less than0.01, multiscale information fusion algorithms simply iterate8times, the algorithm has faster convergence rate. By analyzing the halftone noise spectrum Figure found the halftone noise spectrum is only distributed in the high frequency band, no strong periodic component and the local spectral components from low to high and gradually increase the human visual system is not sensitive.The image of cells extracts the edge effect using canny operator when the best threshold equal to16. According to the image cells of this study engraving process to determine standards through trench values and dark tone value, when the value of the pass ditch and dark tone values in the standard range, the output1of the network points determination the qualified output deemed to have failed. Neural network to establish the quality of gravure network points sorting system, based on the model of the least squares method engraving network points separation results in a the unqualified class, the application of this article half toning method engraving network points of election results are in a qualified class, get the network points pass rates of98%or more.Application of evaluation method based on the quality of the human visual system, the evaluation results can describe the hold of the edge detail, but also the reaction halftone texture suppression level. During the processing of the image halftoning, the results show that the multiscale information fusion method and weighted least squares gradient amplitude and texture entropy, angular second moment were higher than those based on model-based half toning method were lower than model the least squares method. |
PDF Full Text Request