Algorithm Study On The Automatic Detection System For Image Defects Of Color Printer

The industry of printing in China has developed very much in recent years with the progress of science and technology. Especially the gravure shows many advantages such as firm ink layer, wide scope of application, easy maintenance and so on, so it has become the most popular method in printing. However, the printed products sometimes have some defects such as cross-color, ink stain, improper overprint, and scratch etc. Meanwhile, based on the requirements of digital printing and color printing with high quality, the detection system for image defects in color printed matter has emerged accordingly. The traditional defect detection methods such as visual observation, density detection, and chroma detection, cannot accomplish the task online very well due to the subjective factors. With the development of digital video, data acquisition card and computer technique, the application of computer image processing technique in defect detection system for the printed matter becomes feasible. Based on this technique, several companies have already developed some commercial products in China.This work studies the algorithm of automatic defect detection for color printed matter. The work is based on the JH-TCA printed image supervision system. In fact, the hardware of this system aims to accomplish the online supervision, acquisition and store of the printed image with high speed in real-time, and it can compare the actual image with the standard image. However, the noise reduction, image location, image processing and defect identification and detection is not realized. My work studies the image processing algorithm , and accomplishes the experimental design of algorithm on the raw image acquired online. Moreover, the suitable scopes of application for different methods are discussed according to the experimental results.Firstly, the current situation of printing enterprise, the traditional techniques of defect detection, the related research in China and abroad, the discussion on different types of defects in printing, and the basic situations of the printer are introduced in the thesis. Then, in combination with the color mode of printed image, several forms such as RGB, HIS, CMYK are introduced, and their transformation relationships are described. Especially, experimental study is performed to transform a RGB image into a form of CMYK image (four monotonous colors). Meanwhile, the quality of the printed matter is evaluated, which lays a foundation for the later study of defect detection algorithm.Secondly, my main work targets on the design of image sampling module and the development of the signal synchronization unit in the image supervision system. The signal synchronization unit consists of rotary coder, single-chip microcomputer and video. In installation, the rotary coder and the printing plate should be guaranteed to be in line and concentric. The problem of mechanical location for printed image in the procedure of high-speed acquisition with CCD video is well resolved, so the following image processing can be used in the system of image defect detection for color printer. The computer and single-chip microcomputer are combined in this system, which can perform the multiple functions of the mobile equipment, synchronization unit, and image acquisition & store etc. In the procedure of hardware development, the disturbance and analysis of the system noise is discussed. Under the condition that the performance of digital video is improved more significantly while its price is much lower than before, the digital video may be used instead of the traditional analogy camera, in order to reduce some system noise.Thirdly, the fundamental theory and basic methods of image processing are studied. The image grayscale processing and image enhancement processing is performed on the printed image acquired. The suitable conditions for bar graph balance processing and bar graph stretching processing are discussed. The algorithm of image translation and rotation is studied, which makes a good preparation for the later location of image. In the procedure of image location, template matching method is adopted to achieve the image displacement, in order to perform the effective location. The segmentation processing is used for the image to highlight the edge of the image. In the end, the difference image is achieved by using the difference image operator between the actual image and standard image. According to the requirements of the printing precision in detail, different thresholds can be set to judge whether the defects exist or not.Regarding the different types of image processing algorithms discussed in this thesis, Visual C++ 6.0 and Matlab 7.0 is adopted as the software platform. The experimental study is conducted for the defect detection of printed matter. The raw printed image colleted is processed and the experimental result is acceptable, which lays a firm foundation for the realization of automatic defect detection system for the printed image in future.At the same time, with regard to applying digital image processing technique into automatic defect detection for the color printed images, several unresolved problems are proposed here, which are instructive for the future research in this area. The proposals for the future work are given in detail as follows:(1) Multiple arrays of CCD videos should be used in the procedure of image acquisition, in order to resolve the problem in the real-time defect detection of large-size image. The full image acquisition can be accomplished, but the problem is how to stitch the images seamlessly. In the condition of reasonable cost performance, the digital camera should replace the analogy camera gradually to conduct the data collection, which can further reduce the system noise in the sampling process.(2) This work mainly targets on the color printed image. Although the image mode of color printed image is studied and the four-color image is processed uniformly, the problems such as the characteristic information of color image, the overall printing ink, and the ink value in individual printing area still need further study in the future work.