CAJ | 학술논문

针对半色调图像分类中只存在0和1的特点，提出了一种基于改进的协方差矩阵在半色调图像中的分类方法。根据协方差矩阵在实现半色调图像分类中个数少且并未体现其局部和全局信息的特性，对协方差矩阵的底层特征进行改进。利用样本的局部特性和核密度估计方法，实现黎曼流形上的贝叶斯分类策略。实验中研究协方差矩阵的底层特征与传统协方差矩阵的特征提取方法并对其进行分类性能比较。实验结果表明，在半色调图像分类中，与传统的协方差矩阵相比较，改进的协方差矩阵提取出的特征在分类中平均错误分类率更低。
침대반색조도상분류중지존재0화1적특점，제출료일충기우개진적협방차구진재반색조도상중적분류방법。근거협방차구진재실현반색조도상분류중개수소차병미체현기국부화전국신식적특성，대협방차구진적저층특정진행개진。이용양본적국부특성화핵밀도고계방법，실현려만류형상적패협사분류책략。실험중연구협방차구진적저층특정여전통협방차구진적특정제취방법병대기진행분류성능비교。실험결과표명，재반색조도상분류중，여전통적협방차구진상비교，개진적협방차구진제취출적특정재분류중평균착오분류솔경저。
For the characteristic of halftone image classification that only has two values 0 and 1 , this paper proposes an improved covariance matrix classification method based on halftone image. According to the covariance matrix in the realization of a halftone image classification does not reflect the less number and its local and global information on the characteristics , the underlying features of the covariance matrix is improved. Local features of samples and kernel density estimation method are used to achieve Bayesian classification strategy Riemannian manifolds. Experimental study of low-level features of the covariance matrix of the covariance matrix of features with traditional extraction methods to classify performance comparison. Experimental results show that the traditional covariance matrix comparing the halftone image classification and improved covariance matrix extracted feature less in the category average misclassification rate.