Style Translation And Customization For Line Drawings

With the rapid development of computer graphics, the graphics with the goal of non-photorealistic rendering is coming to attach more and more importance. Non-photorealistic rendering (NPR) technique is a brand new and an animate branch of computer graphics. The images generated by NPR often resemble those that, for example, architects, industrial artists, or scientific illustrators produce to communicate more or specific information, often accompanied by text.To accomplish non-photorealistic pictures, researchers have studied and found many effective methods. Style transformation between photo realistic images and non-photorealistic pictures, namely, style learning and transformation has been a hotspot in NPR area. On the side of patterns design and making, some enterprises such as printing, dyeing and ceramics, advertisement, cartoon making, etc, behave poorly in innovation and manufacturing because of difficulty in re-using and adjusting the style. As the multiplicity and complexity of art forms, we could successfully solve the problem if we have a reasonable model to represent the style numerically, to represent the effects of arts with some parameters. Therefore, how to represent the styles formally and control the styles numerically is an important problem.Arts in reality are mostly results of artists' perception, consideration, and abstraction, which relates to the probing into the artistic thinking. There is some difficulty in formally representing and numerically controlling the style. Hence, this paper explored in formally representing and numerically controlling art styles, with the start of simple art form line drawings and extension to models and methods of NPR for other complex art forms.We abstracted style of a line drawing as the styles of its elements: strokes, whichare the trace of the artists' pen. With representation of the strokes' shape and width, we investigated methods of style transformation and customization for line drawings.In this paper, we represent a stroke as{(xi,yi,wi) | 0 < i < n;l;(xo,yo)}, n is the number of control points of the stroke, / is length of it, (xQ, yQ) is the center point Org. The main contributions of this paper are:1. the new method of style transformation and customization for line drawings based on examples. Our method can control the extent of style transformation according to different parameters of process. Then, we gave the customization method. With proportions of every example line drawing, we can make style of the target line drawing have the corresponding similarity to the example. Comparing to previous methods, ours is different in:1) The style transformation can get good results without too much examples.2) The style transformation can be controlled interactively.3) We presented and implemented style customization.2. the new method of style transformation for line drawings based on planar shape evolution. Our idea is evolving strokes of line drawings to get various styles of it based on the shape of a line drawing itself. With the formally representation of the line drawings, this paper presents a more flexible technique to adjust the styles of line drawings. Our novel method has such difference from others:1) The new method can deform the strokes directly based on the evolution theory.2) The direction and amplitude of the deformation can be customized.3) The rules of deformation come from the rules of a line drawing.This paper explored the representation and transformation of the style for line drawings from example learning and direct deformation. Based on these work, we can research the representation and style transformation of other art forms and further implement the transformation from realistic photos to NPR pictures.