彩色图像特征空间变换的新算法及其应用

A New Feature Space Transformation Algorithm in Color Image and Its Application

本文提出了一种彩色图像特征空间变换算法.该算法基于高维形象几何与仿生信息学理论,根据人眼特性,利用彩色图像中彩色信息的变化进行特征空间变换,能够提取出不受光源影响的图像特征.在彩色图像人脸检测应用中,与常见的肤色分割方法不同,该算法不需对肤色建模,不对彩色图像进行非线性彩色空间变换,而是直接将彩色原图变换到特征空间,在特征空间中定位平均脸特征数据的最佳匹配点.本文最后将该算法应用于复杂光源彩色图像的人脸检测中,实验结果验证了算法的有效性和鲁棒性,具有明显的应用价值.

A new feature space transformation algorithm in color image is presented. Basing on high dimensional imaginal geometry and biomimetics informatics and taking advantage of vadations in color information, this new algorithm transforms the color image into feature space and rectifies each feature value within its neighborbood. By using this method, image feature without the effect of different light conditions can be obtained. In face detection application, unlike traditional skin tune detection method, the new method directly transforms the color image into feature space and locates the face position in feature space by face template feature matching. In this paper, the application of this new method to face detection in complex light color images has also been discussed. According to experimental result,it can be seen that the proposed method is very efficient and has significant value in application.