重複圖案是印花織物分色和製版最具關鍵性的組成部分，本研究為因應紡織印花業設計及生產重複圖案管理的需求，創新建立具有重複圖案(Repeated Pattern)及減少影像儲存空間的印花織物影像資料庫，便於圖案檢索。印花圖案包含彩色點狀、條紋、格紋及複雜圖案。
首先於影像預處理，將印花織物彩色樣本透過200 dpi解析度掃描，運用影像灰階化與二維離散小波轉換將影像壓縮成圖像尺寸之四分之一的灰階影像作為分析樣本，接著採用快速傅立葉轉換(Fast Fourier Transform)頻域分析法，將空間域中的印花織物樣本計算每一行與列之頻率振幅圖譜，分析頻譜分佈特性，再藉由倍頻之基頻推算，改善由輸入的非完整週期織物樣本產生之圖形截斷，將所得之基頻利用適應性K均值(K-means)進行分群，群間標準差作為評估分群性能指標，接著分別將行與列最低頻率之群組作為提取頻率之考量，再藉由提取頻率映射回未壓縮前之織物影像，以得印花織物重複圖案。
重複圖案間之相似性分析，可作為判別圖案設計創新性之準則，演算法分為兩階段篩選以減少計算時間及運算量，第一階段將提取後重複圖案與影像資料庫內既有圖案分別以(1)RGB 512色之顏色特徵、(2)影像熵值之紋理特徵與(3)圖案長寬比作為尺度特徵進行相似性分析。餘弦相似性用以評估顏色特徵相似程度，差異百分比用以評估紋理特徵與尺寸特徵間之相似程度。
若第一階段圖案特徵相似性高，則將兩重複圖案進行參數化模板插值匹配，匹配過程中相似性最高區域做為第二階段相似度值，並與所設定創新性之閥值比較，作為存取於影像資料庫之依據。
實驗結果顯示本研究應用快速傅立葉轉換建立印花織物重複圖案影像資料庫之可行性，平均全自動重複圖案提取時間僅10秒/張，所開發之相似性分析統準確率為98.0 %、靈敏度94.4 %為與特異度為98.5 %，適合用於點狀、條紋、格紋及更複雜印花圖案分析，達成資料庫資料共享與生產設計上之需求。

Repeated patterns are the most critical part of the color separation and plate making for printed fabrics. In response to the design and production needs of repeated pattern management in the textile printing industry, this study innovatively created an image database of printed fabrics with re-peated patterns and reduced image storage space for easy pattern retrieval. The print patterns include colored dots, stripes, grids, and complex pat-terns.
First, during image preprocessing, the printed fabric color sample was scanned with 200 dpi resolution. Then, image grayscale and two-dimensional discrete wavelet transform were used to compress the image into a grayscale image 1/4 of the original image size as an analysis sample. The Fast Fourier Transform frequency domain analysis method was adopted to calculate the frequency amplitude map of each row and column of the printed fabric sample in the space domain. The spectral distribution characteristics were analyzed, and the fundamental frequency of the frequency multiplication was calculated to improve the graphics truncation, as generated by the input uncomplete cycle fabric samples. The obtained fundamental frequencies were grouped with adaptive K-means, and the standard deviation between the groups was used as an indicator for evaluating the performance of the grouping. Then, the group with the lowest frequency in the row and column was considered as the extracted frequency, and the repeated printed fabric pattern was obtained by extracting the frequency image back to the uncompressed fabric image.
Similarity analysis between repeated patterns can be used as criterion for identifying the innovation of pattern designs. The algorithm was di-vided into two screening stages to reduce the calculation time and the amount of calculation. In the first stage, the extracted repeated patterns and the existing patterns in the image database were respectively divided into (1) the color feature of RGB 512 colors, (2) the texture feature of the image entropy value, and (3) the pattern aspect ratio was taken as the scale feature for similarity analysis. Cosine similarity was used to evaluate the similarity of color features, and the difference percentage was used to evaluate the degree of similarity between texture features and size features.
If the pattern feature similarity in the first stage was high, then the two repeated patterns were subjected to parameterized template interpo-lation matching. The area with the highest similarity in the matching process was used as the second stage similarity value, which was com-pared with the set innovative threshold, and used as the basis for storing in the image database.
The experimental results showed the application feasibility of Fast Fourier Transform to establish a repeated pattern image database of printed fabrics. The average automatic repeated pattern extraction time was only 10 seconds per sheet. The similarity analysis system, as devel-oped in this study, had an accuracy rate of 98.0%, sensitivity of 94.4%, and specificity of 98.5%. Thus, this system is suitable for the analysis of dots, stripes, and checks, as well as more complex printing patterns, in order to meet the needs of database data sharing and production design.

[1] Kuo CFJ, Lee CL and Shih CY. Image database of printed fabric with repeating dot patterns part (I)–image archiving. Text Res J 2017; 87: 2089-2105.
[2] Huang CC, Liu SC and Yu WH. Woven fabric analysis by image processing: Part I: Identification of weave patterns. Text Res J 2000; 70: 481-485.
[3] Kang TJ, Kim CH and Oh KW. Automatic recognition of fabric weave patterns by digital image analysis. Text Res J 1999; 69: 77-83.
[4] Nishimatsu T. Automatic recognition of the interlacing pattern of woven fabrics. TMSJ 1991; 44: T126-T132.
[5] Kumah C, Raji RK and Pan R. Review of Printed Fabric Pattern Segmentation Analysis and Application. Autex Res J 2019. DOI: 10.2478/aut-2019-0049.
[6] Jing J, Li Q, Li P, et al. A new method of printed fabric image re-trieval based on color moments and gist feature description. Text Res J 2016; 8: 1137-1150.
[7] Oni DI, Ojo JA, Alabi BO, et al. Patterned fabric defect detection and classification (FDDC) techniques: A Review. Int J Eng Res 2018; 9: 1156-1165.
[8] Kuo CFJ, Shih CY, Kao CY, et al. Color and pattern analysis of printed fabric by an unsupervised clustering method. Text Res J 2005; 75: 9-12.
[9] Kuo CFJ, Kao CY and Chiu CH. Integrating a genetic algorithm and a self-organizing map network for an automatically separating color printed fabric system. Text Res J 2009; 79: 1235-1244.
[10] Jiang J, Jin Z, Wang B, et al. A sobel operator combined with patch statistics algorithm for fabric defect detection. KSII T Internet Inf 2020; 14: 687-701.
[11] Wang X, Hou RM, Gao XY, et al. Research on yarn diameter and unevenness based on an adaptive median filter denoising algorithm. Fibres Text East Eur 2020; 28: 36-41.
[12] Schalkoff RJ. Digital image processing and computer vision. New York: Wiley, 1989; 286.
[13] Kuo CFJ, Shih CY and Lee JY. Repeat pattern segmentation of printed fabrics by hough transform method. Text Res J 2005; 75: 779-783.
[14] Mattoccia S, Tombari F and Stefano LD. Fast full-search equivalent template matching by enhanced bounded correlation. IEEE T Image Process. 2008; 17: 528-538.
[15] Kuo CFJ, Shih CY and Lee JY. Separating color and identifying repeat pattern through the automatic computerized analysis system for printed fabrics. J Inf Sci Eng 2008; 24: 453-467.
[16] Zhang J, Zhang H, Wang Y, et al. Automatic Detection of Minimal Repeated Pattern in Printing Fabric Images. In: Proceedings of the 16th ACM Conference on Embedded Networked Sensor Systems, Shenzhen China, 4-7 November 2018, pp. 351-352.
[17] Hu MK. Visual pattern recognition by moment invariants. IEEE Trans Inf Theory 1962; 8: 179-187.
[18] Ullah F and Kaneko SI. Using orientation codes for rota-tion-invariant template matching. Pattern Recognit 2004; 37: 201-209.
[19] Lin YH and Chen CH. (2008). Template matching using the par-ametric template vector with translation, rotation and scale invari-ance. Pattern Recognit 2008; 41: 2413-2421.
[20] Rafsanjani MK, Varzaneh ZA and Chukanlo NE. A survey of hi-erarchical clustering algorithms. Int J Math Comput Sci 2012; 5: 229-240.
[21] Khan S, Nazir S, Hussain A, et al. An efficient JPEG image com-pression based on Haar wavelet transform, discrete cosine transform, and run length encoding techniques for advanced manufacturing processes. Meas Control 2019; 52: 1532-1544.
[22] Cao L, Li H, Zhang Y, et al. Hierarchical method for cataract grading based on retinal images using improved Haar wavelet. Inform Fusion 2020; 53: 196-208.
[23] Tukey JW. Exploratory data analysis, reading, mass. Preliminary edition. Boston, Massachusetts, 1971. Boston, Massachusetts : Addison-Wesley.
[24] Hwang H and Haddad RA. Adaptive median filters: new algo-rithms and results. IEEE T Image Process 1995; 4: 499-502.
[25] Cooley JW and Tukey JW. An algorithm for the machine calcula-tion of complex Fourier series. Math Comput 1965; 19: 297-301.
[26] Kanwal N, Girdhar A, Kaur L, et al. Detection of digital image forgery using fast fourier transform and local features. In: 2019 International Conference on Automation, Computational and Technology Management (ICACTM), Bloomsbury, Lodon, 24-26 April 2019, pp. 262-267. Piscataway, New Jersey: IEEE.
[27] Tanaka K, Sano M, Ohara S, et al. A parametric template method and its application to robust matching. In: Proceedings IEEE Conference on Computer Vision and Pattern Recognition CVPR 2000, Hilton Head Island, South Carolina, 13-15 June 2000, pp. 620-627. Piscataway, New Jersey: IEEE.
[28] Garrido M, Qureshi F, Takala J, et al. Hardware architectures for the fast Fourier transform. In: Handbook of Signal Processing Sys-tems,(eds Bhattacharyya SS, Deprettere EF, Leupers R, et al.), Cham, Switzerland, 2019, pp. 613-647. Cham, Switzerland: Springer.
[29] Zheng X, Lei Q, Yao R, et al. Image segmentation based on adap-tive K-means algorithm. EURASIP J Image Vide 2018; 2018: 68.
[30] Shih CY, Kuo CFJ and Cheng JH. A study of automated color, shape and texture analysis of Tatami embroidery fabrics. Text Res J 2016; 86: 1791-1802.
[31] Xin B, Zhang J, Zhang R, et al. Color texture classification of yarn-dyed woven fabric based on dual-side scanning and co-occurrence matrix. Text Res J 2017; 87: 1883-1895.
[32] Meng S, Wang J, Pan R, et al. Recognition of the layout of colored yarns in yarn-dyed fabrics. Text Res J 2020. DOI: 0040517520932830.
[33] Zhang Z, Bai S, Xu GS, et al. Research on the knitting needle de-tection system of a hosiery machine based on machine vision. Text Res J 2020. DOI: 0040517519899173.
[34] Hu MK. Visual pattern recognition by moment invariants. IEEE Trans Inf Theory 1962; 8: 179-187.
[35] Swain MJ and Ballard DH. Color indexing. Int J Comput Vis 1991; 7: 11-32.
[36] Tanaka K, Sano M, Ohara S, et al. A parametric template method and its application to robust matching. In: Proceedings IEEE Conference on Computer Vision and Pattern Recognition CVPR 2000, Hilton Head Island, South Carolina, 13-15 June 2000, pp. 620-627. Piscataway, New Jersey: IEEE.