The Design And Implementation Of Human Face Pencil Portrait Simulation System Based On Non-photorealistic Rendering

The topic described in the thesis makes full use of the latest achievements in computer vision, machine learning and image processing, and demonstrates the design and implementation of the human face pencil portrait system based on non-photorealistic rendering.The process of the design is as follows: Firstly, we employ Mean Shift algorithm and Graph Cut algorithm to segment input images with a small amount of interaction from user. The natural background, the human hair, the human face and neck and the cloths are all separated. Secondly, we utilize Active Shape Model (ASM) search algorithm to detect human face and ENT precisely. Thirdly, based on the theory of primal sketch and the result of detecting human face, we generate the oriental field of the image, which extracts the texture information of the image, especially the facial region. Then we make white noises of the image. Last but not least, we select the Line Integration Convolution algorithm for producing the texture effect and the lighting effect grounded on the oriental field and white noises. Both of the two effects are mixed together to generate vivid human face pencil portrait effect.The interface implementation and the running effects of the system are also be displayed and analyzed. It can be proved that the final effects of the system simulation are successful and ornamentally artistic.