Texture Classification Based On Local Directional Patterns

In the field of image analysis,texture classification is one of the most important research topics and local directional pattern (LDP) is a typical texture operator.In recent years,researchers have proposed a variety of extensions of LDP.Based on the analysis of LDP and its extensions,the improved schemes are proposed in the paper.The main contributions of this dissertation are summarized as follows:(1)A novel method is proposed based on fuzzy theory.Firstly,the fuzzy logic is introduced to overcome the rigid classification of the original directional patterns and the membership function is adopted for flexible partition.Then,the traditional local structure is transformed from 3×3 neighborhood to multiscale region.It is more robust to image noise.In addition,the mean fuzzy LDP algorithm is proposed by taking the mean value as the fuzzy threshold.The results show that the proposed schemes perform better than the traditional methods.(2)In order to achieve noise robustness for the traditional LDN,the normalization and average filter are introduced to reduce the influence of image noise.In addition,an improved LDN algorithm is proposed to fuse the difference information between the neighborhoods.The new scheme encodes the convolution direction information and the intensity information for texture classification.The experimental results show that the new algorithm can greatly improve the performance of the original methods.(3)On the analysis of local derivative pattern (LD_eP) and local directional gradient pattern (LDGP),a novel operator,MDP (Multi-directional Gradient Pattern) is proposed in the paper.The new operator integrates the different directional derivative information and the directional features of neighborhood pixels together.The experimental results show that the new algorithm achieves higher scores than the methods mentioned in the paper.