基于PCA矢量形态学的彩色图像分割方法

Color Image Segmentation Method Based on PCA Vector Morphology

基于HSV彩色模型,提出一种基于主量成分分析的矢量排序新方法。将彩色图像描述为矢量空间,则像素点作为矢量处理,根据相应彩色矢量对主量轴的投影值进行排序。采用这种统计特征排序方案,定义新的上下确界与彩色形态算子,将其应用于彩色图像分割,能得到与视觉判断相一致的分割结果。实验表明,该算法具有出色的矢量保持能力,与标准的彩色形态学算子相比,显示了更优的分割性能与较好鲁棒性能。

This paper presents a new approach to vector ordering based on Principal Component Analysis（PCA） in HSV color model. Color image is represented by a vector field and each pixel is imposed on color vector, color vectors are ordered according to the projection scores obtain by projecting each color vector on the principal axis. Based on the vector ordering scheme, supremum and infimum are defined, and the color morphological operators are extracted. By applying the vector order statistics to color image segmentation, result holds favorable consistency in terms of human perception. Furthermore, the experiment demonstrates the algorithm is vector preserving, compared with standard color morphological operators, has better segmented properties and robust.