Research On Non-photorealistic Rendering Techniques Based On Image Segmentation

Non-photorealistic rendering of computer graphics is one of the most active research topic, it is not like realistic graphics that pursue just like a picture's reality, but hope that it can highlight the image to express information through ignore some of the unimportant details, or to generate the image with artistic effect. As the artist paint different regions with different properties of drawing strokes, based on the study of non-photorealistic rendering technology, this article study the research of pencil drawing and painting simulation based on image segmentation.This article's work includes: study the development of non-photorealistic rendering and image segmentation algorithm, and analyze some key technology of non-photorealistic rendering, including the contour diction technique, rendering technology based brush model, the non-photorealistic rendering based on texture synthesis and color transfer algorithm; study of several image segmentation algorithm of image processing and describe in detail the efficient segmentation algorithm based on graph theory than used in this paper. Finally, study the technology of pencil drawing and painting and propose a method of pencil drawing and painting simulation based image segmentation.In the process of pencil drawing, this paper used a method based on the gradient sharpening to detect the contour of picture. Transform the original image to gray-scale and make a segmentation, on this basis, give the statistics of average gray value of each region, and add different white noise to each region using different algorithm. In the segmentation, based on the characteristics of each area of information extraction, produce the vector of each field, then use the LIC algorithm to get the texture image, and fuse the contour map and texture image to get the picture with the pencil drawing effect. The simulation of oil painting, according to the characteristics of oil painting-when artist painting they usually use different stokes to paint in different regions or paint different objects. So, the image can be drawn in each region, and in each area or according to the objects that want to paint, select different brush with different attributes.The innovation of this paper is: first, the image contour extraction, a sharpening of contour lines based on the gradient extraction method, the algorithm first gradient image sharpening, and then reversed to color processing can get the image contour information. Second, on the image to use the efficient method based on graph partitioning, the algorithm was improved. After the image is converted to an undirected graph, to judge and figure the weight of each edge of the information, when the weight of edge is less than the threshold that have set as T, the edges connecting two vertices merge as an area to complete an initial segmentation of the image, and then use the efficient segmentation method based on graph theory to segment, the segmentation algorithm to improve efficiency. Third, using the improved graph-based segmentation algorithm for image segmentation and then according to the different characteristics of each region to set the attributes of the brush, each segment is filled with this brush. In the drawing process, strokes are terminated when they reach a maximum length or reach the boundary of the region.