中文摘要：

提出一种新的描述纹理的方法——视点切割（Viewpoint Slicing）模式。方法依据视觉心理学的理论，提取人的视觉系统对之敏感的纹理信息，如纹理的全局灰度极值点，局部灰度极值点，基元的边缘等特征。方法可以描述基于统计的方法无法适用的基元比较大的纹理，也可以有效描述复杂的自然纹理。详细讨论了方法的有效性，并用提取的纹理特征进行纹理分类实验，在Brodatz纹理全集上达到了96.7％的高正确分类率。而且方法基本不用调整参数，实验证明不同的参数配置都能稳定地达到相当高的正确分类率。

英文摘要：

We propose in this paper a novel texture description method through Viewpoint Slicing（VPS）. It reflects the psychophysieal evidence that the human visual system is sensitive to global and local gray level extremums and edge information. VPS can describe textures consisting of large primitives which most classical statistical methods can not work well with. It can also describe complicated natural textures well. We discuss the effectiveness of VPS in details and use the extracted features for texture classification. Experiment results show that VPS achieves a high correct classification rate of 96.7% on the entire Brodatz set. There is little need to tune the parameters because different combinations of parameters can achieve similarly high correct classification rates.