Automatic genaration of sketch-like pencil drawing from image

This paper proposes an automatic scheme for generating sketch-like pencil drawing from natural images. The proposed method consists of two main steps: a stroke emulating step to depict obvious outlines and a tone generation step for filling textures. In the stroke emulating process, a hierarchical edge detection method is designed to extract line segments with different resolutions, which simulates object contour drawn by artists using multiple strokes in different level of details. In the tone generation step, an extended ETF direction field is constructed firstly, then a parameterized swing bilateral LIC filter is applied to generate pencil texture in appropriate tone and direction. Finally, the pencil texture is transferred by simulating physical effect of pencil drawn on paper. Experiment results show the pencil drawings generated by the proposed method exhibit high similarity compare to the work drawn by artists.