摘要
提出了一种基于红外焦平面阵列读出结构的非均匀校正算法,它将校正过程分成两步。首先利用恒定统计方法滤除由通道放大器引起的固定图案噪声,然后采用改进的神经网络方法,进一步校正由探测器响应产生的非均匀性。实验结果表明:该算法在第20帧时就能够收敛到一个较好的水平,此刻校正图像的平均绝对误差(MAE)和峰值信噪比(PSNR)值分别为2.9和39.2。由于本算法是递推进行的,并且具有收敛速度快、计算复杂度低等优点,因此易于工程实现。
A nonuniformity correction algorithm based on infrared focal plane array readout architecture is proposed to correct the aggregate nonuniformity by two separate steps. Firstly, the nonuniformity from the readout amplifiers is corrected using the constant statistics method. Then, the noise resuited from the detectors is decreased with the improved neural network method. The experimental results indicate that the proposed algorithm can converge to a good correction level at the frame 20, while the Mean Absolute Error(MAE) and Peak Signal-to-Noise Ratio(PSNR) of the corrected image are 2.9 and 39. 2, respectively. The algorithm is recursive with the advantages of high convergent speed and low computational complexity,so it is easy to be implemented in projects.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2008年第1期128-133,共6页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.60572151)
关键词
红外焦平面阵列
非均匀校正
固定图案噪声
恒定统计
神经网络
infrared focal plane array
nonuniformity correction
fixed-pattern noise
constant statistics
neural network