紋理分類技術在電腦視覺的應用上，扮演相當重要的角色，在過去幾年儘管已有許多這方面技術的提出，但克服因環境變動如旋轉及缩放所造成紋理分類不一致的現象，仍是最主要的課題。 基於「直覺模糊集合」（intuitionistic fuzzy sets, IFSs）我們提出嶄新的整體與局部特徵，此兩個特徵能描述3個像點間的微紋理結構及其統計資訊，這整體特徵稱為「模糊樣式直方圖」（fuzzy motif histogram, FMH），而局部特徵稱為「模糊樣式頻譜」（fuzzy motif spectrum, FMS）；同時，我們也設計出系統架構與直覺模糊集合的相似度比對，透過實驗證明，我們提出的方法不僅有高準確率，並且對旋轉及缩放亦具強韌度。 此外，經實證發現，一些最新的紋理分類方法在利用直覺模糊集合為基礎之下，均較原來提昇準確率。最後，我們的方法也應用在彩色雷射印表機列印來源的鑑識研究，除證明極佳的辨識效果，並建議未來建立印表機列印紋理專屬資料庫的可行性，以提供資料鑑識與犯罪偵查的使用。 ;Texture classification plays an important role in computer vision and has a wide variety of applications. Many methods of color texture analysis have been developed over the years; however, a major problem is that textures in the real world are often not uniform owing to variations in rotation and scale. In this thesis, we propose novel features at both the global and local levels—namely, the fuzzy motif histogram (FMH) and the fuzzy motif spectrum (FMS)—using statistics and microtexture information spread across three pixels (i.e., higher-order statistics). Thus, this method achieves texture classification based on the intuitionistic fuzzy sets (IFSs) theory. Furthermore, we offer a system framework and a similarity measure between two IFSs. By conducting many experiments, we explored the effectiveness of the proposed methods, as well as their robustness against image changes, such as changes in rotation and scale. Additionally, it was found empirically that for texture classification, several state-of-the-art IFS-based methods always achieve higher accuracy than non-IFS-based methods. Finally, we used our proposed system for color laser print identification and conducted a feasibility study to examine the system’s potential for use in digitizing a subject-specific, laser print database as part of data forensics and crime investigation operations.