对MOS电阻阵列非均匀性产生的原因进行了深入分析，并提出了一种可明显提高红外图像生成质量的 新的电阻阵列非均匀性校正方法。通过离线测试得到整个电阻阵列的电压–温度数据和待校正单元的电 压–温度数据。以这些原始数据作为神经网络的输入，建立电阻阵列中每个单元的校正前输入和期望输 入之间的函数关系，生成查找表，从而解决MOS电阻阵列由各单元响应率不同而引起的非均匀性问题。仿真结果表明， 该方法具有可行性。与国内外常用的分段线性化方法相比，其校正效果有了很大的提高。

The causes resulting in the nonuniformity in a MOS resistor array are analyzed and a new nonuniformity correction method which can obviously improve the quality of the generated infrared scene is proposed. When the MOS resistor array is tested offline, the voltage vs temperature data and the voltage vs temperature data to be corrected are measured. Then, by taking these raw data as the input of a neural network, the functional relationship between the pre-correction input and the expected input of each element in the resistor array is established. Finally, a look-up compensation table which can be used to reduce the nonuniformity due to the different responsibility of each element in the MOS resistor array is generated. The simulation result shows that this method is feasible and its correction effectiveness is greatly improved in comparison with the conventional piecewise linearization method.