基于BP网络的图像增强器像元通道不一致性校正研究

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。

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

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.