CNI评估指标

CNI评估标准

1、Transforming color image in RGB space to other color space such as CIELUV color space.

将RGB空间中的彩色图像转换为其他颜色空间，如CIELUV颜色空间。

2、Computing the luminance (L), hue (H), and saturation (S), respectively.

分别计算亮度（L）、色调（H）和饱和度（S）。

3、Thresholding L and S components: L values between 20 and 80 are kept, S values over 0.1 are kept.

阈值L和S分量：L值保持在20和80之间，S值保持在0.1以上。

4、Defining three kinds of pixels according to hue value: 25–70 is called ‘‘skin’’ pixels, 95–135 is called ‘‘grass’’ pixels, and 185–260 is called ‘‘sky’’ pixels.

根据色调值定义三种像素：25-70称为“皮肤”像素，95-135称为“草”像素，185-260称为“天空”像素。

5、Calculating averaged saturation values for ‘‘skin’’ $S_{average\_skin}$, ‘‘grass’’ $S_{average\_grass}$, and ‘‘sky’’ $S_{average\_sky}$ pixels and numbers of ‘‘skin’’ pixels $n_{skin}$, ‘‘grass’’ pixels $n_{grass}$, and ‘‘sky’’ pixels $n_{sky}$.

计算“skin”$S_{average\_skin}$、“grass”$S_{average\_grass}$和“sky”$S_{average\_sky}$像素的平均饱和度值，以及“skin“像素$n_{skin}$、“grass’”像素$n_{grass}$和“sky”像素$n_{sky}$的数量。

6、Calculating local CNI values for ‘‘skin,’’ ‘‘grass,’’ and ‘‘sky’’ pixels:

计算“skin”、“grass”和“sky”像素的CNI值：
$$\begin{gathered} N_{skin}=exp(-0.5\ast ((S_{average\_skin}-0.76)/0.52{)}^2) \\ {N}_{grass}=exp(-0.5\ast ((S_{average\_grass}-0.81)/0.53{)}^2) \\ {N}_{sky}=exp(-0.5\ast ((S_{average\_sky}-0.43)/0.22{)}^2) \end{gathered}$$
7、Calculating the global CNI values:

计算全局CNI值：
$$N_{image}=(n_{skin}\ast N_{skin}+n_{grass}\ast N_{grass}+n_{sky}\ast N_{sky})/(n_{skin}+n_{grass}+n_{sky})$$
注：CIELUV：LUV色彩空间

参考文献：Huang K Q, Wang Q, Wu Z Y. Natural color image enhancement and evaluation algorithm based on human visual system[J]. Computer Vision and Image Understanding, 2006, 103(1): 52-63.