一种基于主矢量集的纹理图象分析方法

A Method Based On Principle Vector Set For Texture Analysis

描述了一种基于多维纹理主矢量集与相应能量集的纹理图象表示及分析方法。并给出了由纹理图象求得主矢量集及相应能量集的计算过程。为了将该方法用于纹理识别及分析，文中讨论了在上述表示时，纹理图象的相似度度量，给出结构误差及能量误差的计算公式及两种计算结构误差、能量误差及总误差的算法。最后给出了用上述方法由计算机对Ｂｒｏｄａｔｚ的自然纹理图象及人工合成的规则纹理图象分析的实验结果，并与共发矩阵法作了比较。

This paper describes a texture representation and an analysis method based on multi dimensional texture principle vector set(PVS) and its corresponding energy set (ES), and also provides an approach to computing the PVS and the ES from a texture image . For the convenience of texture analysis and recognition, we present an algorithm based on the PVS and the ES, with which a structure error, a energy error and a total error between two texture images are able to be obtained. The algorithms have been implemented on PC platform to evaluate the performance of the method, also have given two experiment results and the comparison with the co-occurrence matrix method. The results of the experiments demonstrate that the method is efficient and robust.