Shading-aware Non-photorealistic Rendering Of Human Face Based On Photo

Non-photorealistic Rendering (NPR) of human face has wide application in many aspects such as cartoon making. How to render faces of various styles with three-dimensional, vivid effect is a difficulty. Shading can strengthen our understanding to the scene depth, make the scene seems more vivid, so it plays an important role in the NPR for face.In this paper, we present a framework of shading-aware NPR of human face based on photo, which can easily change the shading effect of the cartoon face. Given a facial photo, we first detect the contour, wrinkle and other important features from the input image, then, generate a facial cartoon. Meanwhile, we decompose the input image into shading image and reflectance image. After that, we pose non-linear enhancement or thresholding on the shading image, then, add the edited shading image to the vectorized image to generate the final result. With the decomposition, our algorithm can edit on the shading image to get various result with different shading. Experiments demonstrate the flexibility of our algorithm to edit the shading of the image, which make the render result more vivid. Our algorithm has great potential in application.