Automatic Synthesis of Oil-Style Paintings

• Main Authors: Bo-Hui Wu, 吳柏褘
• Other Authors: 黃于飛 Ph. D.
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/61158631049698212686

碩士 === 國立宜蘭大學 === 資訊工程學系碩士班 === 102 === Due to the rapid advances in camera technology, it is much easier and more convenient to capture photos for any user. There is more and more image editing software available, but the existing software that simulates oil-painting effect either requires manual assistance or needs specific combination of different filtering functions. Therefore, this thesis aims to develop an easy-to-use oil-painting synthesis software program. Non-photorealistic rendering (NPR) is an important topic in the research field of Computer Graphics. In this thesis, the goal is to design a fully automatic approach, not targeting on real-time application. The proposed algorithm takes a single image as the input and applies NPR technology to synthesize the oil-painting effect. Different from the existing approaches that use real brush stroke images as templates, we create our own brush strokes in random manner to enhance the variety and manipulation of the strokes. For determining the direction of the brush, we also proposed a new method based on template-matching to evaluate the direction of the image edge features. Many image processing algorithms only consider eight directions, but our method is able to distinguish 16 different directions. Our algorithm first paints the color-homogenous regions, then the edge regions, and finally fills up the remaining pixels. The experiments show that synthesis time is between 50–200 seconds for a 600x800 images, and the simulation results are appreciated.