基于多特征融合的彩色印刷图像识别算法

Image Recognition of Color Printing Based on Multi-Feature Integration

针对当前彩色印刷图像的识别技术及设备存在的不足,提出一种改进的Shearlet变换和颜色特征分析的图像检索技术,以实现对某幅印刷图像进行精准识别。通过对当前主流彩色图像识别技术的研究,以Shearlet变换、颜色直方图和支持向量机(SVM)理论为基础,逐步完成图像边缘提取、图像颜色特征提取和图像边缘与颜色特征融合的图像识别过程。通过研究和相关实验表明,Shearlet模极大值确定和自动最佳阈值选取是图像边缘检测的关键,HSV颜色直方图的分解、量化、合成是图像颜色特征的有效途径,SVM分类器的分类功能设计是多特征融合识别实现的手段。在精确获取图像边缘轮廓和颜色信息的基础上,所提出的多特征融合图像识别算法,可以将识别误差控制在10%以内,并为后续印刷图像质量控制提供强有力的技术支撑。

In order to overcome the disadvantages of current technology of color printing image recognition and indentify a printed image more precise?ly,proposes a new image retrieval technology based on Shearlet transform and color feature analysis.Current mainstream technology of col?or image recognition mainly consists of Shearlet transform,color histogram and support vector machine(SVM),which are employed in our paper to accomplish image edge extraction,color feature extraction,and the fusion of image edge extraction and color feature extraction.Shows that the Shearlet modulus maxima and the automatic selection of threshold are significant to image edge detection.Also shows that the decomposition,quantization,synthesis of HSV color histogram is an effective way to extract image color feature and the classification function of SVM classifier is a useful method to realize multiple features fusion.Experiments validate that the proposed image recognition algorithm and the recognition error can be limited within10%,which can provide stronger technical support for subsequent printing image quality control.