Research On Image Based Rendering Of Chinese Ink Painting

Non-photorealistic rendering (NPR) is a research field combining computertechnology and painting art. It refers to generate non-photorealistic, with hand-drawnstyle graphics using computer. Chinese ink painting is a unique rendering artisticstyle of China, which has obvious difference with western artistic on drawing toolsand painting techniques. Using computer to generate art works have Chinese inkpainting style, which can make full use of computer technology and expand the fieldsof computer applications, promote national culture, and it has important theoreticaland practical significance.Current Non-photorealistic rendering techniques can be divided into twocategories according to the different date source: one is based on three-dimensionalobservation space, which use3D scene geometry model as input, the other istwo-dimensional artistic rendering based on image space, which use an original2Ddigital image as input. Most simulation methods use3D model, and renderingprocedure is complicated. In this paper we combine the digital image processingtechnology with computer graphic methods, take2D image as input which is easy toobtain, and can be used in a wider range.In this paper, we studied the main simulation methods of Chinese ink Painting, andanalysis the characteristics of Chinese ink painting. On this abasement we proposedan image based rendering method of Chinese painting, which can transform a2Dinput image into an image has Chinese painting style automatically. We describedthis algorithm in details.Through observing and studying the procedure of artist rendering Chinese painting,we can find, the artist often use fewer brush strokes to rendering the outline of the rendering objects, and use color blocks to color the interior. Inspired by the realpainting process in Chinese painting, we segment the input image into differenttarget objects, and for each object, the whole rendering process is divided into threemain steps: take use of the brush model we established to depict the outline of theobject, for the interior of the object, we process the color information in the colorspace, and finally we build a paper model to simulate the ink diffusion effect.In the procedure of stroke rendering, we adopt a method which combines thethreshold image segmentation method and region growing method, taking intoaccount the spatial information and color information, segment the input image intodifferent drawing objects, and find the borderline of each object. Using a mechanismto generate successive strokes have different length. Considering the width change ofreal strokes in Chinese painting, we set different width value to each control pointand obtain the strokes we want to rendering.In the modeling of brush, we do not simulate the physical characters as mosttraditional methods, and still not calculate the deformation of brush during themoving process, because our system is not an interactive system and do not need tosimulate the brush effect in screen. Instead, we want to simulate the effects ofChinese ink painting, so we consider the shape of the contact area of the brush andthe paper. We adopt ellipse to simulate the contact area, and the bristles aredistributed in it. According to the change of center point during the renderingprocedure, compute the location of the bristles, and fill in the drawing strokes tosimulate the strokes as ink painting.As for the internal color process, because most artists often use chunks of ink tocolor the internal of the objects, and the gray levels can be detected by human arelimited, so we transform the input image into gray image and divide the interior areainto five gray level using a histogram equalization method, to make the input imagehas less gray scales and high contrast, resulting the effect similar to the large shadedarea in Chinese painting. Now we synthesize the drawing results through the two steps above, and havegenerate image has Chinese ink painting effect. But the rendering effect is not natural,there is not good transition between the edge and interior, so we establish a papertexture model, for each paper cell, generate the fiber number it contains randomly,and calculate water absorption. If the pixel ink amount in the image has a gap withthe neighboring pixels, or the quantity of ink exceed the amount of ink the paper cellcan accommodate, the ink will proliferate to the neighboring pixels, the method isrelatively simple, to some extent, it can simulate the effect of ink diffusion inChinese ink painting, and can better reflect the flavor of Chinese painting.In this paper, an algorithm which can generate a Chinese Painting from an inputimage is proposed. It take2D image as input, it's a combination of image-basedrendering and non-photorealistic technology. Without the intervention of the user,and the user do not need to have prior knowledge and drawing skills, it can realizethe stylize painting. In this algorithm, we establish the brush and paper model, as forthe modeling of brush, this method focus on the simulation of rendering effect, ratherthan the analog to the physical characteristics of the drawing medias, therebyreducing the amount of computation and complexity of algorithm, the paintingefficiency is greatly improved. Through the modeling of paper, we simulate the inkdiffusion effect, make the rendering result has the flavor of Chinese ink painting.Experimental results show that the algorithm can simulate the characteristics ofChinese ink painting to some extent, and it is feasible in theory and practice.