基于图像块先验的单帧红外自适应校正算法

摘要

The nonuniformity of the infrared focal plane array severely limits the sensitivity of the infrared imaging sys-tem and reduces the imaging quality of the image. An expected patch log likelihood( EPLL) algorithm based on the statistical characteristics of image patches was introduced,and a single frame image nonuniformity correction algorithm based on image patches priori was proposed,which can accomplish the nonuniformity correction of single frame infra-red image. Firstly,the Gaussian mixture model is used to train and classify the image patches,and then the EPLL algo-rithm is used to obtain the corresponding maximum log likelihood probability of the infrared image with nonuniformity noise. The experimental results of the simulation images and the real infrared images show that the algorithm has good effect on the nonuniformity correction of single frame infrared image,and it has the fast calibration speed,and can a-void the generation of ghost imaging.%红外焦平面阵列的非均匀性严重限制了红外成像系统对目标的灵敏度,降低了图像的成像质量.本文基于图像块统计特性的期望块对数似然概率(EPLL),提出了基于图像块先验的单帧红外图像非均匀性校正算法,可以完成单帧红外图像的非均匀性校正.该算法首先利用高斯混合模型完成对图像块训练分类,然后利用EPLL准则获取带有非均匀性噪声图像对应的最大似然概率对数期望图像来完成对图像的校正.仿真试验和真实红外图像实验结果表明该方法对单帧红外图像的非均匀性校正效果良好,校正速度快,可以避免鬼影的产生并且可以完成持续性工作.