基于亮度的自适应色度空间模型肤色算法研究

Research on Self-adaptive Chroma Space Model Skin-color Algorithm Based on Brightness

肤色检测是人脸识别和敏感图像过滤等领域中的关键技术。针对已有肤色算法对亮度信息考虑不足的缺点,在色度空间模型基础上提出了一种在彩色图像中检测肤色像素的新方法——基于亮度的自适应色度空间模型肤色算法,该方法综合了肤色像素在YUV、YIQ色彩空间上的分布特征,并且可以根据图像亮度信息以及I值和UV的夹角值区分肌肤颜色,针对不同肤色选用不同阈值进行检测。实验证明该方法对偏黄、黑色肌肤(亚洲、非洲人)和偏白色肌肤(欧美人)均有很好的检测效果,肤色像素检测的正确率达93.4795%,误检率和检测时间相对变化不大,检测效果有明显提高。

Skin-color detection is the key technique of face recognition and erotic image Filtering etc. Aim at ex-isting skin color algorithms considering the brightness information not enough,a new approach of detecting skin-color pixels in multicolor images based on chroma space model——self-adaptive chroma space model skin-coloralgorithm based on brightness is presented,the approach colligates the distributed feature in YUV and YIQ colorspace,and differentiates skin color according to the brightness and angle of UV of images. It chooses differentscopes corresponding to different skin-color.Experiments show obvious superiority on yellow,black (Asian,African) and white skins (Occidental).The correctness of skin-color detection is improved to 93.4795%,littlechange of error rate and detection time,detection result is improved obviously.