肤色纹理保留实时人脸美化算法

摘要

In order to overcome these deficiencies, such as the information loss in the skin texture, eyes and hair areas, a novel beautification algorithm based on skin texture preserving is proposed. Firstly, the skin-color- simi-larity map is computed by adopting the adaptive skin-color model on chrominance component. Secondly, the skin roughness map is estimated based on the local variance of the image, and then the smoothing-control parameters are adjusted adaptively for calculating the texture-preserving image. Thirdly, the skin-color-similarity map is used to fuse the original image with the skin texture-preserving image; Finally, the fusion image is whitened to obtain the final beautified image. Extensive experiments performed on the public face database demonstrate that the proposed algorithm can effectively eliminate the skin blemishes or flaws, and retain the visual-sensible skin de-tails simultaneously. Compared with the state-of-the-art methods, proposed algorithm can achieve better beauti-fying effect, as well as with lower computation complexity.%针对目前人脸美化算法容易产生皮肤纹理信息损失,眼睛和头发等区域信息丢失等缺陷,提出一种具有肤色纹理保留性能实时人脸美化算法.首先根据自适应肤色椭圆模型处理色度分量估计肤色相似度图;然后对原图局部方差处理得到肤色粗糙度,据此调整肤色纹理保留平滑参数,得到肤色纹理保留平滑图像;再利用肤色相似度图融合原图与肤色纹理保留平滑图像;最后对融合图像进行美白处理,获得人脸美化图像.在公用人脸数据库上的测试结果表明,文中算法不仅有效地祛除皮肤瑕疵,还能保留人眼敏感的肤质细节,使视觉感知比较自然;与现有美化算法对比,该算法美化效果更佳,鲁棒性强,复杂度低,能实时处理人脸视频.