本論文介紹一個處理紋路分析的新方法，叫做紋路特徵編碼法。它結合灰階明亮度相互關係矩陣及紋路頻譜的優點。根據紋路特徵編碼法所產生的紋路特徵數統計圖及紋路特徵數相互關係矩陣可以導出許多紋路特徵。在實驗中我們將這個方法與灰階明亮度相互關係矩陣、紋路頻譜，和交叉－對角線紋路矩陣等三種紋路分析方法進行比較，以了解紋路特徵編碼法在分辨紋路影像方面的效能。從實驗結果可知紋路特徵編碼法比較其他三種紋路分析方法，有較佳的正確性且較不易受影像旋轉的影響。 This paper introduces a new texture analysis method named Texture Feature Coding Method (TFCM). This method incorporates with the merits of both the gray-level co-occurrence matrix (GLCM) and texture spectrum (TS) methods. The texture feature number histogram and the texture feature number co-occurrence matrix are generated by TFCM for derived many useful texture feature descriptors. Three texture analysis methods, GLCM, TS, and cross-diagonal texture matrix (CDTM), are used to compare and evaluate the performance with TFCM in discriminating some of Brodatz’s natural textures. The experimental results reveal that the TFCM is superior to other three methods in classification of natural textures, especially in classification of rotated textures.
