Single infrared stripe nonuniformity correction algorithm based on adaptive diffusion models
-
摘要: 针对红外焦平面成像系统存在列向条纹非均匀性的现象,采用了一种基于自适应PM扩散模型的非均匀校正新算法。首先,综合利用图像梯度信息和局部灰度统计信息,自适应计算PM模型的扩散阈值;然后将每列像素的PM模型估计值作为该列像素的期望值;最后采用最陡下降法迭代计算得到每列像元的校正参数,并对结果进行循环校正以提高校正效果。实验结果表明:该算法可以保护图像边缘信息,与同类算法相比,能够更有效地抑制条纹非均匀性,并且能够防止图像产生鬼影。Abstract: In order to correct the stripe nonuniformity for infrared images captured by infraed focal plane array(IRFPA), a novel stripe nonuniformity correction algorithm based on adaptive PM diffusion models for single infrared image is adopted.Firstly, the adaptive diffusion threshold of PM model is calculated by gradient information and local gray level statistics of infrared images.Then, the estimate values of each column pixel are treated as expectations, which are in constraint of PM models.Finally, correction parameters in iteration are obtained by method of steepest descent, and the image is corrected repeatedly to improve correction performance.Experimental results indicate that the adopted algorithm can preserve edge information.Compared with other four algorithms, the proposed algorithm has advantage of reducing stripe nonuniformity and removing ghosting artifact.
-
图 1 不同扩散阈值对应PDE-based NUC校正前后的图像
Figure 1. Images before and after using PDE-based NUC algorithm correction with different thresholds
图 2 PDE-based NUC算法中对图像27列偏移校正系数和误差函数的估计
Figure 2. Estimations of the offset coefficients and the error signal using the PDE-based NUC algorithm
图 3 基于自适应扩散阈值的非均匀性校正算法流程
Figure 3. Block diagram of the NUC algorithm based the proliferation of adaptive threshold
图 4 5种算法用于实拍外红外图像的单帧校正效果比较
Figure 4. Comparison of five NUC algorithms for signal infrared real images
表 1 5种算法的平均校正效果比较
Table 1. Average correction performance comparison of five NUC algorithms
Performance NN-NUC MMSE LSK-SK PDE-based NUC PM-AD-NUC RMSE 24.670 3 24.749 1 18.580 5 18.510 2 16.248 9 ρ 0.165 6 0.167 1 0.152 5 0.157 5 0.144 9 RMSE(var) 1.784 2 1.846 1 1.057 5 1.006 5 0.728 4 ρ(var) 4.845 2×10-4 9.031 7×10-4 2.088 7×10-4 2.842 7×10-4 1.989 2×10-4 Time/s 5.88 3.76 4.09 4.71 7.45 
下载: 导出CSV
-
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] -
下载:
点击查看大图
计量
- 文章访问数: 1010
- HTML全文浏览量: 110
- PDF下载量: 907
- 被引次数: 0
