Depth-based Anisotropic Kuwahara Filtering
Jeongho Ahn, Jong‐Chul Yoon, In‐Kwon Lee · 2010
When artists draw a picture of photorealistic scene in an image, they describe only specific parts that represent characteristic features carefully, but they express the parts about less important region roughly. In study about Non-photorealistic rendering, image abstraction research reflects such artist's character. Thus, methods about image abstraction commonly preserve image features and flatten non-feature area. Recently, Kyprianidis et al. [2009] introduced Anisotropic Kuwahara Filtering (AKF) which generates feature preserved image abstraction using the smoothed structure tensor. However since they used only color information to defining anisotropic ratio, different regions that have similar color are conquered by each other unintentionally. Hence, we propose the depth-based AKF method that considers not only color, but also depth to generate image abstraction where boundary feateures are effectively preserved.