基于BP神经网络的阴极射线管显示器标定模型

Display calibration model of cathode ray tube based on BP neural network

为了精确标定阴极射线管(CRT)显示器,应用分光光谱辐射计1980B采集计算机控制的CRT显示器色度数据,建立了经过三通道输出整合优化的前馈BP神经网络CRT显示器标定模型.该模型是CRT显示器的红、绿、蓝3个通道的亮度等级值和颜色三刺激值之间的非线性映射.神经网络的预测颜色三刺激值与测量颜色三刺激值之间的决定系数均大于0.999 9,两者之间的CIE色差均值为0.409 7,标准差为0.332 3,远小于最小颜色可觉差.该模型的标定色差显著小于CIE伽马校正标定模型的标定色差(CIE色差均值为4.695 5,标准差为3.533 0).

Spectroradiometer 1980B was used to collect the chromaticity data of computer controlled cathode ray tube （CRT） display for precise calibration. Based on optimized feed-forward back-propagation （BP） neural network through a three output channels integration, a CRT display calibration model was provided to establish the nonlinear relationship between the intensity scale values of red, green, blue channels and the tristimulate values. The determination coefficient of neural network predicted tristimulate values and measured tristimulate values were both greater than 0. 999 9. The mean CIE color difference between the predicted and the measured tristimulate values was 0. 409 7 with standard deviation of 0. 332 3, which was far less than the minimum perceptible color difference （MPCD）. Compared with the mean color difference （mean value was 4. 695 5, standard deviation was 3. 533 0 ） of CIE gamma correction model, the color difference of the BP neural network model was significantly smaller.