A new system for detecting fabric defects using image processing based on the pattern classification by Likelihood function was proposed. The system is made up of a sensing module, a preprocessing module, a feature selection module, a categorization module, and a pattern classification module. The sensing module obtains the digitized image of the fabric using a scanner. The categorization module divides a fabric into defect or detect free area using K-menas algorithm. By using K-means algorithm, the new system is more useful than any other previous systems because of its low memory size and high speed. The pattern classification module recognizes the defects based on the Basian classifier. The decision function and decision boundary are effective identifiers because they reduce time consumption and abridge the previous procedures. A training plain weave fabric containing several slubs and neps is used to find the decision function. The developed system using this decision function detects defects in a test plain weave fabric, and then records a defect report. A defect position report showed reasonable results and high percentage of correcting detection.