A crack detection system was developed to observe growth behavior of fatigue crack using the digital image processing techniques. The crack propagation behavior was observed successfully by combining block matching method and inclination threshold value method. The data obtained with this new system agreed with the existent data and the standard deviation was 0.03%. If the proposed method is utilized to develop a crack monitoring system to observe the crack growth behavior automatically, the time and effort needed for a fatigue test could be dramatically reduced. And also it will be easier to estimate and to manage efficiently the safety and residual life of industrial facilities.