A Stylistic Method Using Extended DOG Operator

NPR(Non-photorealistic rendering) is a branch of computer graphics and is different from photorealistic rendering, more and more researchers begin to focus on NPR, NPR is a kind of method to produce stylistic abstraction of a photograph. The edge detection technology have a lot of applications in NPR field. Experimental and theory proved that advanced image stylization with extended Difference-of-Gaussians(DOG) has giant potential in stylistic rendering. Thus, the main research is extended method based on DOG.The Difference-of-Gaussian (DOG) operator can be used to emphasize or detect edges which achieved significant effect. Recent the research of extensions to the standard Difference-of-Gaussian edge detection operator be put more on its aesthetic appeal for its stylistic application and less susceptible to noise, while omit the technical subtleties of the DOG operator. This paper put forward a kind of stylistic method which control several parameters since the parameters of extended DOG are easy to control. This method defined an ETF of input image, then begin to filter on the base of ETF. A lot of new results spanning a variety of styles are achieved, including pencil shading, pastel, negative edges and anti-aliasing. The experiment represents that this formulation is easy to achieve, the visual quality is comparable to or better than signal stylistic system.This stylistic method using extended DOG operator is belong to edge detection technology, thus this paper introduce several simple edge detection methods, and devise math model of extended DOG operator in part two. Part three of this paper introduce three main problem, they are constructing ETF, filtering on base of ETF and making LIC. The forth part of this paper include algorithm steps and several stylistic results, these results are negative edges, anti-aliasing and controlling threshold.