Design And Implementation Of Color Image Separation System

With the development of the CAD/CAM, in order to improve the competitiveness in the market the Chinese traditional previous technics processing has changed to automatization from the method in workshop. However, because the the image was scanned from the cloth and it is impacted by the spin texure and it is distortion by the dye absorption, which result in the less precision of the Color Separation Algorithm. Hence at present most apply the manual scanned draft to the Color Separation but it increase the manual editing workload and it is inefficient. It is conflict to the fast delivery and high quality in the market. Therefore it is required a method to improve the precision of the Color Separation. In our paper we have studied this problem and it has acquired high value in application.In this paper, firstly considering local and global, color and spatial and edge information, we proposed a novel algorithm of textile image separation based on graph cut. The novel algorithm is composed of four steps. First, we input strokes to get the main colors according to the observation of a user in the regions where the colors looks like one color. Then, choose one of colors as the foreground and the rest colors as background in turn, and use the graph cut technology to obtain every segmentation region. Also we delete the small and narrow regions. Third, we compute the probability of boundary of the textile image, and consider the color, spatial and edge information to label the left pixels. Lastly, we use the average color of each input color region as the rendering colors to instead the old image, so we can get the image separation result. This algorithm acquires preferable chromatography images and improves the precision of the color separation and avoids the impact from the noisesform the original image.The image was scanned from the cloth and it is not only impacted by the noise from the cloth, but also by the dye absorption. It can cause shape deformation. This can increase editing workload, so we need to modify the pattern after image color separation.Amendment to the pattern is divided into two parts, one part is to amend the simple geometry directly, another is to search similar pattern from the pattern database to amend. In order to amend the simple geometry, we compute the Centric Moment of each object in the pattern respectively, and then determine their categories which belong to an oval, triangle, quadrangle or protruding polygon that has more than four edges. After this, we use specific approach to amend the geometry under different categories.For the ungeometrical patterns, we use pattern matching approach. In order to search the proper pattern, the pattern database needs to be established. The color separation algorithm described in the paper is mended to construct the pattern database. Two strokes are inputed by the user, one expresses foreground and another expresses background. The color separation algorithm operates to divide the pattern's pixels into two groups, and the foreground pixels are extracted to save into the pattern database. When matching, the needed patterns are queried from the database according to the shape features, and the shape feature vector is calculated through the Fourier Descriptor. Firstly, using canny operator to obtain the edge image, and then accessing to their profile images, and then using Fourier descriptors to compute the shape feature vector. Compare Reed Euclid distance of this shape feature vector to vectors in the pattern library. Return a list of selection for users in descending order with the similarity of the pattern match. Choose the most similar pattern to replace the original patterns to achieve the purpose of the amendment.There are two innovations in the paper. One is that the proposed color separation algorithm adopts a variety of image features and produces accurate result. And the algorithm can be used to extract the pattern. Another is that after color separation, the pattern amendment is provided to reduce the workload of manual editing and improve the work efficiency.The system designed in the paper provides solutions for the issues in actual applications. Firstly the color separation is executed, and then the resulting pattern is amended to achieve better effect, which meets the demand of rapid production...