One of the few manual operations in the almost fully automated manufacturing of LCD(Liquid Crystal Displays) is visual inspection process. Along with the need of multifold increase of production it becomes currently central to automate visual inspection process. The following defects must be found : Pin Holes, White Spots, Black Spots and other flaws larger than those of allowable size. The main difficulty of automating the LCD inspection problem is known to be the detection of little defects in real time.
This paper describes a flexible image processor - IPSH(Image Processor with Soft Hardware) with preprocessing algorithms including Moving Average filtering & Window Subtraction Method in order to automate LCD visual inspection process. The hardwarized preprocessing algorithm which is to extract image features and unidirectional ring structure of DSPs (TMS320C40) render the suggested image processor a time effective solution to this kinds of inspection problems. Finally, validity of suggested image processor is shown via simulation and experiments.
