一种新的基于直接最小二乘椭圆拟合的肤色检测方法

肤色检测是计算机视觉中的一个重要问题,本文提出了一种新的基于直接最小二乘椭圆拟合的肤色检测方法,其基本思想是根据肤色样本分布区域的边界数据点采用曲线拟合的方法得到肤色分布区域的边界方程。在实现时,为了解决直接在笛卡儿坐标系中提取肤色样本分布区域边界数据的困难,算法采用了一种新的解决思路,即首先把训练肤色样本在色度空间的统计分布转化为图像的形式,然后再利用边缘检测方法得到肤色分布区域的边界数据。根据所得的边界数据点用直接最小二乘椭圆拟合方法便可得到肤色分布区域的椭圆边界,方法简单直观。实践表明,该算法能完成对各种不同环境条件下所拍摄图像的肤色分割,效果理想,其性能明显优于常用的域值界定法和单高斯模型法。

A New Skin Color Detection Algorithm Based on Direct Least Square Ellipse Fitting

Skin color detection is an important problem in computer vision.A new skin color detection algorithm based on direct least square ellipse fitting is propesed in this paper.The basic idea of the proposed algorithm is to attain the skin color distribution boundary equation via curve fitting.In order to solve the difficulty of acquiring the boundary data of skin color distribution region in the Cartesian coordinate,a new method is used,that is the statistical distribution of skin color training samples in chrominance space is transformed into the image form firstly.Then the data on the edge of the skin color distribution region are extracted via edge detection.Fi- nally,the skin color distribution ellipse boundary can be attained using direct least square ellipse fitting.The proposed algorithm proves to be simple and intuitive,experimental results on images taken under a wide range of different environment demonstrate that it is effi- cient and superior to the method using explicit skin region boundary and single Gaussian model.