Compensation of motion blur in binary images

Image processing is used very often for industrial applications. The most common input devices are CCD-cameras. Blurring in the image can arise from moving and swinging objects. This motion blur should be compensated for estimation of geometrical quantities. An interrelation for the shape of opposite edges in the image is used for the reconstruction of the binary image. In many cases a simple threshold decision can be applied. The paper describes an algorithm for threshold adaptation using a self-learning system. The self-learning process occurs in the background and does not influence the real time processing. Some experimental results illustrate the performance of the proposed method