基于BP网络的图像增强器像元通道不一致性校正研究

Study on nonuniformity correction of XRII-CCD pixel channel response based on BP neural network

像元通道响应不一致性是影响图像增强器检测能力的重要因素,因而有必要对不一致性进行校正。首先分析了传统的不一致性校正方法的缺点,然后提出了一种基于BP神经网络的像元通道响应不一致性的校正方法。该方法利用BP神经网络良好的非线性映射能力,对每一个像元通道进行输入-输出特性的反非线性逼近,同时利用实验数据进行了神经网络的训练。对比了校正前后的图像及标准差。

The pixel channel response nonuniformity is a major factor affecting the detection ability of the XRII-CCD. Therefore it is necessary to develop correction algorithms to compensate for the nonuniformity. The disadvantages of the traditional nonuniformity correction methods are analyzed firstly. Then a correction approach of XRII-CCD pixel channel response based on BP neural network is proposed. The nonlinear reflecting capability of BP neural network is applied to approach the I/ O characteristic curve of each pixel channel. And at the same time , neural network is trained by the I/O data measured. The images and the standard deviations for the images are given before and after the correction. The nonuniformity correction capability of the presented solution is validated by experimental results.