Abstract: This study describes an automated computer vision system for plain ceramic wall tile surface inspection. The system is designed to detect surface cracks on the inspected tile at its final stage of production and to categorize its into either accept or reject class based on the complexity of the occurring and detected cracks. Hardware implementation for the system consists of a monochrome CCD area scan camera with a 512�512 resolution, ring-light illuminating unit and an IBM-PC compatible Pentium IV MMX 2.4 GHz computer. Our approach to cracks detection involves the application of image edge enhancement and detection algorithms to the captured digitized image of the inspected tile using Prewitt masks and the development of efficient image thresholding algorithms to segment out the cracks from the image background. Some geometric features useful for tile classification were selected and measured from the detected cracks. Classification algorithms were developed by training and testing the network with enough number of the measured features from all defective tile samples stored in the database by means of iterative, learning algorithms. Simulations and experiments conducted on real tile images show the correctness and efficiency of the developed algorithms computed and implemented in this study in real-time.