摘要： 对现有的动态聚类算法进行改进，提出一种Lab颜色空间中基于兴趣点动态聚类分析的颜色分级方法。在考虑视觉监测实时性和计算准确性的基础上，通过色适应变换和对比敏感度函数滤波，补偿人眼视觉系统的空间混合效果，采用基于兴趣点的动态聚类分析提取颜色特征，根据视觉容差、彩度和色度的依赖关系，确定色差度量方法，采用最小分类器进行颜色分级。实验结果表明，该方法的平均色差仅为2.36，分类计算的时间范围为500 ms~700 ms。

关键词: 机器视觉, Lab模型, 颜色分级, 颜色聚类, 色彩量化

Abstract: This paper improves the existing dynamic clustering algorithm, and presents a color grading method based on interest points and dynamic clustering analysis. In addition to considering the real-time and accuracy of visual monitoring system, color data is compensated for the effect of spatial scale of the human visual system by color adaptability transform and a spatial filter with Contrast Sensitivity Function(CSF). Then, color features are extracted based on interest points and dynamic clustering analysis. According to the dependencies among in visual tolerance, chromes and chromaticity, color difference measurement is determined and the minimum distance classifier is adopted to color grading. Experimental results show that the average color difference is only 2.36, classification calculation time in between 500 ms and 700 ms.

Key words: machine vision, Lab model, color grading, color clustering, color quantization

中图分类号: 

• TP391