With the increase of people's demand for the quality of presswork, traditional methods which detect presswork's defects with human eyes is checking the development of printing industry. In order to improve presswork's productivity, print enterprise have to look for new technology in automation field. Machine vision has been applied in presswork's detecting, which greatly improve the speed and the accuracy of detecting, and become new projects of research and application by presswork enterprise.Machine vision system mainly simulates human vision through computer. It process and understand information from objects' image to detect, control, survey. It can get information quickly and then process immediately, so it is widely used in automatic production.The experiment design a machine vision system to detect presswork's defects. First of all, this paper introduce the background and significance of presswork's defection and point out machine vision technology is a good scheme for the problem. Secondly, it give us working principle of the machine vision system to design and the characters of several great components. Algorithm research is an important aspect in design works. After analyzing three typical algorithms, introduce a new algorithm and experimental result on it, which base on dynamic threshold. Finally, a software of defect detection system for presswork is developed. We can operate the defect detection system through the software and see real-time processing result on man-machine interface.The defect detection system for presswork built in this paper meets the real-time requirement. The system function is relatively complete and convenient to be used.