Color Print Registration Method Based On Pattern Recognition

Color image segmentation and target detection are important topics in areas of computer vision and pattern recognition, and have significant value in studying pattern recognition and computer vision technology. Color-printing-registration is used to distinguish each registering marks of printing colors form the color printing sample image.In this paper, aimed at the color printing sample image sequences which got under the unstable working environment, we go deep study into some effective methods of image segmentation and target detection. And according to the features of color space in color printing, and the characteristic of the shape of the registering mark, some ameliorations has been proposed in this paper:1. According to the features of the color space for color printing color, transform the color image from RGB color space to CYMK color space at first, and then separate the image by the ameliorated separating method. First convert the color printing image into serious gray image with only one color component for color printing, and then using image segmentation method to separate the registering marks form every gray image, lastly using the same image segmentation method to separate the objective registering mark form all registering we separated from gray image. By this method we coule separate the objective registering mark form the color printing sample image more accurately.2. According to the charactristic of the registering mark which is used for color printing registration, we extracting the skeleton form the separated image at first, and then use the quickly Hough transform algorithm to detect the registering mark from the skeleton. That is, firstly extract the skeleton of the objective marks, and then use the quickly Hough transform algorithm to detect the shape and location charactristic of the registering marks from skeleton we extracted. By this method we could detect the objective registering mark form the separated image with much less time and does not lose the precition of the classic Hough transform algorithm.In this paper, this method has been applied by some color printing images which collected form the camera for printer; the result shows that this method is feasible.