【作者】 陈雨；

【导师】 胡涛；

【作者基本信息】 西安理工大学 ， 信号与信息处理， 2004， 硕士

【摘要】 针对当前印刷领域中印刷图像色彩处理方式的稳定性及存在的不足，本课题提出了一种新的基于数字图像处理和模式识别技术的色彩处理方式，即印刷色彩空间转换及分色原理，同时设计出了一个彩色印品显微网点图像分色系统。分色系统借助CCD数码相机和光学显微镜等设备，获取彩色印品显微网点图像的RGB的色彩数据信息；而后采用数字图像处理和模式识别技术对网点图像中各像素进行颜色模式识别，最终得到各单色网点图像的分色结果以及各色网点面积率。分色系统中算法模型的设计充分体现了当今人工智能方面的前沿技术，如神经网络，信息融合等。具体构建了为便于识别的两个分类知识库，即初分类知识库和细分类知识库，这些分类器的设计综合了模糊逻辑，聚类算法、神经网络、以及信息融合等方面的技术应用，使分色系统具有识别精度比较高及速度较快的优点。课题对各种分类器的构建过程及工作原理进行了详尽的阐述，对分色结果进行了总结归纳和误差分析，同时指出了后续研究方向。更多还原

【Abstract】 According to the drawbacks and stability of algorithms in color processing of printing images in current printing industry, a new method for transforming color spaces and color separation based on digital image processing and pattern recognition is presented, and a system of color printing micro-dot image has been designed . With CCD digital camera and optical microscope, color printing micro-dot image RGB data information can be got, then adopt the digital image processing and pattern recognition techniques to recognize the color patterns of pixels in the micro-dot images, at the same time, the color separation and area coverage can be retrieved.The design of algorithms model fully embodies the advanced techniques in the field of AI, such as neural network, information fusion. In the color separation system, two classifying knowledge libraries have been established, namely, one is the rough knowledge library, and the other is the fine knowledge library. The two classifying libraries have comprehensively applied the fuzzy logical, clustering algorithms, neural network and information fusion, made the classifiers have the merits of high degree recognition and fast speed. The task has introduced the construction procedure and principals of the classifiers in details, analyzed the errors and summed up the color separation results. In the end, it also indicated the direction of study in the future.更多还原