Abstract: A new method for color face recognition is proposed in order to make full use of color information but with little complexity.First of all，it extracts S and V components in HSV color space，and then extracts the spectroface and FLD feature separately，finally，it classifies and recognizes the feature according to the fusion of the two components.Extraction of only S and V can both retain the color information and reduce the complexity.The spectroface is actually the amplitude of Fourier transform of the wavelet general image，and its shift invariance and rotation invariance can effectively eliminate the error caused by the expression change and a little mask.Experiment on the ESSEX color face database verifies that this method has a good ability to recognition.

Key words: color face recognition, color space, spectroface, Fisher linear discriminant analysis, feature fusion

摘要： 为了充分利用彩色信息而又不使算法过于复杂，提出了一种对彩色人脸图像进行识别的新方法，首先基于HSV空间提取S、V分量，然后分别采用频谱脸与FLD对其进行特征提取，再以两分量的融合特征作为分类依据进行识别。选取典型特征S和V，既保留了彩色信息又降低了复杂度。频谱脸实际上就是小波概貌图像的傅里叶幅度谱，其位移不变性和旋转不变性能有效消除因为人像表情变化和少许掩膜带来的识别误差。通过对ESSEX大学彩色人脸库的识别实验，验证了该方法有很好的识别能力。

关键词: 彩色人脸识别, 颜色空间, 频谱脸, Fisher线性判别分析, 特征融合