Abstract: Now, the registration precision usually adjusted by handwork that is an undeveloped method for multi-color printing machines in china. To solve this problem, an automatic detection method of color registtration is proposed. Registration marker image is photographed on the basis of visual imaging principle, and the calculation method of registration deviation is established by adopting the image analysis theory. Marker images are preprocessed by using the vector signal filtering method of color image, and image objects are extracted, so a color recognition model for cyan, magenta, yellow and black primary colors and their complex color pixels is derived. The next step is that edge pixels of color marker are extracted, and their structural lines are approximated by least square fitting and Hough transform method. At last, relative distances among registration markers on image are calculated and translated into relative distances in object space via vision calibration methods. Experiments indicate that precision of 20 micron is obtained when this automatic detection method is applied. And this accords with general printing technology standard of china. This method can be applied to two or four color offset printing system based on the CMYK synthesis principle.

Key words: Hough transform, Image analysis, Least square fitting, Registration