A Study On Pencil Drawing Generation Based On Hierarchical Drawing

As a branch of computer graphics,Non-photorealistic Rendering has been more than 20 years of history.Corresponding to Realistic Rendering,Non-realistic Rendering techniques do not require the computer to draw seemingly real scenes and objects,but rather to abstract the important features of the rendered scene,producing images with a particular style or application,such as the simulation of artworks of different styles or the visualization of scientific data,and thus derived out many application software.Since getting attention at 1990s,the field of Non-realistic Rendering has produced a number of quite mature methods to simulate specific art-style images,such as simulation of pencil drawing,oil painting,printmaking,ink painting and so on.Pencil drawing synthesis as an important topic in the field of non-realistic drawing has gained the attention and research of many scholars.According to the input data,these pencil synthesis methods mainly divided into two classes:(1)the three-dimensional model-based methods and(2)two-dimensional image-based methods.This thesis presents an automatic scheme for generating pencil drawing from natural full-color images based on hierarchical drawing.The proposed method consists of two main steps:a stroke emulation step for depicting obvious contours and a texture generation step for filling shading information.The stroke emulation process designs a hierarchical edge detection procedure to extract line segments in different resolutions,which simulates the process of drawing contours using multiple strokes as done by artists.The texture generation step constructs an extended ETF direction field from the input image,and applies a parameterized swing bilateral LIC filter to generate pencil texture in appropriate tone and direction.Finally,the pencil texture is further improved by simulating physical effects of pencil sketching on paper.Experiments are conducted to generate pencil drawings from natural images with vary contents,and the results show the pencil drawings generated in the proposed method exhibit high similarity with hand-drawn version compare to previous methods reported in the literature.