关键词: Mean Shift图像分割, FDoG边界提取, 卡通化渲染

Abstract: Non-smooth extracted boundaries or lack of salient region segmentation are major drawbacks of existing image cartoon style rendering algorithms. To deal with this problem, a cartoon style rendering method based on Mean Shift and FDoG is presented. Given an image, the method acquires a cartoon object with salient segmentation of regions and smoothly connected boundaries from the image by using techniques of both region segmentation and edge extraction, combined with the luminance quantization strategy and other post-processing technologies for cleansing objects to finally generate a cartoon style image. More specifically, the method first segments the image via Mean-shift algorithm into valid regions by setting appropriate parameters, and then adopts FDoG algorithm to extract image edges, and finally gets the final image with cartoon style by removing or merging visually disturbing regions and by quantizing the image’s luminance channel with reference to cartoon image’s luminance distribution. The proposed method is easy to implement and the satisfactory experimental results is gotten.

Key words: Mean Shift image segmentation, FDoG edge extraction, cartoon style rendering