基于Matlab神经网络工具箱的颜色模型转换方法研究

由于计算机颜色量化系统的每个设备都具有不同的色域和色彩特征，颜色信息在该系统中的精确控制及传递就显得尤为困难，所以采用彩色亮度计对颜色量化系统进行标定。摄像机获取的彩色图像被表示成RGB值，从彩色亮度计获得XYZ值，两者之间的转换是非线性的，因此引入神经网络完成这一转换。借助MATLAB神经网络工具箱完成了训练程序、测试程序、预测程序的编写工作，用经过归一化的黄色训练样本与测试样本对隐层节点数为12的三层网络进行训练，结果表明经过训练的BP网络可以满足RGB空间向CIEXYZ颜色空间转换要求，达到了预定目标，使用该网络可以定量表达颜色。

Study on conversion method of color space based on Matlab neural network toolbox

Since each equipment of computer color quantification system has different color gamut and feature, accurate control and transmission of color information in this system is particularly difficult. Therefore the col- or luminance meter is selected for marking the color quantification system. The color picture captured from CCD camera is expressed by RGB values and XYZ values are gotten from the color luminance meter, so the neural network is introduced because of the non-linear conversion. The training program, testing program and foreseeable program is completed respectively by Neural Network Toolbox in Matlab applications. The three-lay- er network with 12 nodes in the hidden layer is trained based on the processed normalization yellow data of training set and testing set. The results show that the network by training can satisfy the conversion between RGB space and CIEXYZ space, which has reached the scheduled goal. This network can express the color quantitatively.