A pencil drawing algorithm based on wavelet transform multiscale

Non-photorealistic rendering (NPR) plays an essential role in the field of graphics, which has been pursuing the simulation and reproduction of artistic styles. In recent years, more and more researchers have been attracted to process the natural images into pencil drawings. Many pencil drawing algorithms have been developed. But the existing pencil drawing algorithms including the problems of missing picture details and coursing edge lines cannot get the desired results. Therefore, a novel pencil drawing algorithm generated by multi-scale system is proposed, which is based on the artist's hand-drawn pencil drawing process. Firstly, wavelet transform is utilized to generate the multi-level contours of pencil line which reserve more detailed informationand the edge of lines can be more delicate and soft. Next, histogram matching is used to obtain the same tone as the artist's hand drawn pencil drawing. The same texture as the artist's hand-drawn pencil drawing is obtained by simple splicing small pieces of texture. Then, an optimized combination approach of the tone and texture is adopted to obtain tonal texture. Finally, pencil drawing effect map with abundant picture information and smooth lines is obtained by merging contours of pencil line and tonal textures by pixel. By applying the proposed algorithm to all kinds of natural images, the experimental results show that the pencil drawing obtained by proposed algorithm is much closer to the artist's hand-drawn pencil drawing. In addition, the proposed method improves the efficiency of image processing and performs well in both reserving detailed information and expressing the original scene vividly.