工業自動化的時代裡，在高效率與高品質的市場需求下，如何能使產品又快又好的生產出來是許多工業技術改良的目的，因此生產流程中部分依靠人力完成的工作逐漸的被機械設備所取代。 為了確保品質，複雜的生產線中會設下許多的檢測點，檢驗每一個程序是否有瑕疵品的產生，而這些檢測的工作早期多仰賴人工費時費力的檢查。 近年來機械視覺的技術逐漸被應用於自動化檢測的系統，如何快速且正確的檢測出瑕疵存在的位置是許多研究的主題。
檢測瑕疵是一種同中求異的概念，在大部分相類似群體中找出相異的個體。 同樣的，在本文討論的瑕疵是存在影像相類似背景訊息中特徵歧異的區域，因此檢測目的即是為了找出這類區域。 一般常見瑕疵檢測演算法可以正確的檢測瑕疵的存在，卻無法得知瑕疵實際的位置，因此本論文以紋理分析理論為基礎，提出一種以鄰近灰階相依矩陣為核心的瑕疵定位方法，能夠正確且迅速的檢測出瑕疵，並標定出其確切的位置。
紋理的種類可以分為隨機、結構及決定性紋理三種，本論文以結構性紋理做為檢測實驗的對象，除了驗證一般瑕疵的檢出正確率外，更是著重於具有偏轉現象之特殊紋理的實驗與探討。 紋理分析中灰階共生矩陣與離散餘弦轉換一直是常用且頗受好評的方法，在本文中將把自行發展的瑕疵定位演算法與這兩種方法做一番的比較與討論。

In the era of industry automation, machine equipments have replaced traditional manpower in many production processes for better efficiency. To ensure the quality, if the production process is complicated, many checkpoints are generally established such that defects can be detected and damage can be reduced, or even prevented completely. In the early days, human inspection, which not only is labor intensive but also time consuming, is the main tool in quality control. With the improvements in machine vision techniques, machines have gradually taken over this task from human again. As a result, a lot of research efforts have been devoted in improving the accuracy and the speed of defect detection in recent years.
In nature, the defect detection is to find the black sheep in a bunch of supposedly identical objects. In this research, we are trying to locate a special region which has some different characteristics in a background which has some identifiable pattern throughout the target image. Unlike the conventional defect detection techniques which generally only tell the target is defect or not, our new research is trying to reveal the location of the defect additionally. Based on a Neighboring Gray Level Dependence Matrix, we are able to develop a new technique which can achieve the aforementioned goals in a texture image.
In general, texture can be classed into random, structural, and deterministic based on its nature. In our research, we are concentrated on the detection and localization of the defects in a structural texture. Special efforts have been devoted on reducing the possible misfire when the image, and hence the texture, is rotated. Comparing to the conventional techniques, our techniques has much better detect rates with much less computational costs.

[1] Leen-Kiat Soh, “Texture Analysis of SAR Sea Ice Imagery Using Gray Level Co-Occurrence Matrices”, IEEE Transactions on Geoscience and Remote Sensing, Volume 37, Issue 2, Part 1, March 1999
[2] G. Lambert, F. Bock , “Wavelet Method for Texture Defect Detection”IEEE Image Processing, Volume 3, Oct. 1997
[3] Ajay Kumar , “Defect Detection in Textured Materials Using Optimized Filters”, IEEE Transaction on Systems, Man, and Cybernetics, VOL, 32 , NO. 5, OCT. 2002
[4] Chi-ho Chen, Granthan K. H. Pang. “Fabric Defect Detection by Fourier Analysis”, IEEE Transactions on Industry Application, VOLUME32, NO.5, 2000
[5] Ajay Kumar ,”Defect Detection in Textured Materials Using Gabor
Filters” IEEE Transactions on Industry Application, VOLUME 38, NO.2, 2002
[6] Gosta H. Granlund , Hans Knutsson “Signal Processing for Computer Vision” 1995
[7] L. H. Siew, R. M. Hodgson, and E. J. Wood,” Texture measure for carpet wear assessment“ IEEE Tran. Pattern . Anal. Machine, Intelligence ,
VOLUME 10, NO. 1, Jan 1988
[8] Fumiaki ,Saburo ”Computer Analysis of Visual Textures” 1990
[9] Tamura, H., S. Mori, and Y. Yamawaki, “Textural Features Corresponding to Visual Perception. ”, IEEE Transactions on Systems, Man, and Cybernetics, 1978
[10] Sklansky ,J., “Image Segmentation and Feature Extraction”, IEEE Transaction on Systems, Man, and Cybernetics,1978.
[11] Haralick, R.M., “Statistical and structural Approaches to Texture ”, Proceedings of the IEEE, VOLUME 67, NO. 5, MAY 1979
[12] Mari Partio, Bargdan Cramariuc “Rock Texture Retrieval Using Gray Level Co-Occurrence Matrix”, Proceedings of the 5th Nordic Signal Processing Symposium, 2002
[13] Ahmet Latif Amet, Aysin Ertüzün, Aytül Erçil “Texture Defect Detection Using Subband Domain Co-Occurrence Matrices” , IEEE Southwest Symposium,1998

[14] Brodatz, P., Texture: A Photographic Album for Artist and Designers. New York, Dover Publications, 1966
[15] Shuttleworth, Bennett, “Colour Texture Analysis Using Co-occurrence Matrices for Classification of Colon Cancer Images”, Proceedings of the IEEE, 2002
[16] Jon R. Berry, John Goutsias “A Comparative Study of Matrix Measures for Maximum Likelihood Texture Classification”, IEEE Transaction on
Systems, Man, and Cybernetics, VOLUME 21, NO. 1, JAN, 1991
[17] Sang-Mi Lee,Hee-Jung, ”Efficient Content-Based Image Retrieval Method Using Color and Texture ”ETRI Journal volume 20,1998
[18] K.R. Rao, J.J. Hwang , “Techniques and Standards for Image, Video,
Audio Coding” Prentice-Hall Inc. New Jersey U.S.A 1996
[19] 陳同孝、張真誠、黃國峰，數位影像處理技術，松崗，2001
[20] 黃文吉，C++ Builder與影像處理，儒林，2003
[21] Borland C++ builder 討論區http://delphi.ktop.com.tw/default.asp
[22] http://www.ux.his.no/~tranden/brodatz.html