利用质心模式统计进行图像纹理分析

基于区域质心具有某些与人类视觉相适应的性质,提出一种新的纹理分析算法——质心模式统计法(StatisticsofCentroidPattern,SCP)。该算法通过对纹理基元质心分布位置与特点进行分析得到纹理的基元特征,该特征可以反映基元内灰度、能量等分布信息。然后通过对基元特征进行加权投影统计,得到图像的方向性、对比度等纹理特征,这些特征可以更好的适应人类视觉特性。将该算法应用于Brodatz图像库的图像检索,实验结果表明,SCP在查准率上比差分矩阵和纹理谱分别提高10.26%和26.66%,取得了更好的检索效果,可以对图像纹理进行更为精确的描述与分析。

Image texture analysis by statistics of centroid patterns

A new texture analysis algorithm?Statistics of Centroid Pattern (SCP) is proposed on the basis that the position of region centroid can match the human vision. In the algorithm, the features of texture primitives are obtained by analyzing centroid position and distribution, which represent the distribution of gray and energy. The texture features such as direction and contrast of the image can be obtained from weighted projection statistics of primitive feature, which are more matched with human vision. To test the performance of SCP, we take a CBIR using Brodatz set. Experimental results show that the SCP is 10.26% and 26.66% higher than that of differential matrix and texture spectrum in precision, respectively, which prove that SCP can depict and analyze textures better.