摘要
差分矩阵算法在对图像纹理粗细程度进行识别时,只采用视觉感知的方式,难以区分分布形态非常相似的差分直方图.为此,采用描述性统计量峰度和偏度给出差分直方图的数值解释,建立基于差分直方图的纹理定量描述算法.对brodatz自然纹理图像进行的实验验证了算法的有效性.此外,对于一组给定图像,该算法可以对其按照纹理粗细程度进行判别和排序,并通过选择合适阈值对纹理图像进行分类.
The differential matrix algorithm used the way of visual perception only to identify the coarseness-neness degree of the texture in the image,and the distributions of differential histograms with similarity could not be distinguished.In view of this problem,the numerical explanation of differential histogram using descriptive statistics ＇kurtosis and skewness＇ was proposed,and the quantitative description algorithm of texture based on differential histogram was established.The verification experiments were implemented using brodatz natural texture image database,which supported the validity of the proposed algorithm.Moreover,for a given group of images,the proposed algorithm can be used to distinguish the images by the coarseness-neness degree of the texture,and sort the images by choosing suitable threshold.
出处
《大连海事大学学报》
CAS
CSCD
北大核心
2011年第4期71-74,共4页
Journal of Dalian Maritime University
关键词
差分直方图
纹理分析
峰度
偏度
differential histogram
texture analysis
kurtosis
skewness
