基于改进局部方向模式的纹理分类

针对局部方向模式(Local Directional Patterns,LDP)及其扩展方法存在的问题,提出了一种增强的局部方向模式方法。首先,针对传统LDP及其改进的刚性模式划分策略,基于模糊逻辑理论,通过引入模糊隶属度函数来提高模式划分的准确性。其次,对传统的局部3×3邻域进行了扩展,新的拓扑结构不但可实现多分辨率分析,而且进一步降低了噪声的影响。采用在纹理分类领域广泛应用的UIUC、Curet和Outex纹理图像库进行试验,结果表明新的方法可以显著提高纹理图像的分类效能。

Texture Classification Based on Improved Local Directional Patterns

In order to improve the performance of the traditional local directional patterns (LDP) and its extensions,an improved LDP method is proposed.Aiming at the traditional LDP and its improved rigid pattern partitioning strategy,based on the fuzzy logic theory,the accuracy of pattern partitioning is improved by introducing the fuzzy membership function.Then,the traditional local structure 3×3 of the local directional patterns is extended to multiscale region.The new topology can not only achieve multi-resolution analysis,but also further reduce the impact of noise.The widely used texture databases,such as UIUC,Curet and Outex,are used as test beds,and the results show that the proposed schemes perform better than the traditional methods.