为提高不同光照条件下的人脸识别精度,提出了一种利用多级小波分解和样条插值进行光照补偿的方法。一般认为,光照变化会给图像中混入两种噪声:背景噪声和增益噪声。背景噪声是加性噪声,可以通过直接对图像进行多级小波分解和样条插值来估计并去除;增益噪声是乘性噪声,可以对去除了背景噪声的图像进行对数变换后,再利用多级小波分解和样条插值来估计并去除。在Yale B人脸数据库上的实验结果表明:本方法的识别率比其它方法高2%～11%,可以有效地对不同光照条件下的人脸图像进行调整,提高不同光照条件下人脸识别的精度和稳定性。

The appearance of face will vary dramatically when the illumination changes,which makes face recognition even more challengeable.In this paper,an illumination compensation method by means of multi-level wavelet decomposition and spline interpolation is proposed to reduce the effect of illumination variation on face recognition.The light variation in an image can be modeled as additive noise and multiplicative noise.The additive noise,called background,can be estimated and eliminated by multi-level wavelet decomposition followed by spline interpolation;the multiplicative one,called gain,can be done in the similar way in the logarithm domain.Experimental results on Yale B face database show that the proposed method achieves 2%～11% recognition rate,higher than that of the others.It can effectively adjust the facial images under varying illumination conditions,as well as improve the recognition rate and robustness of face recognition.