基于混合自适应回归的红外盲元补偿算法

摘要

An infrared blind-pixel compensation algorithm is proposed based on hybrid autoregressive modeling (HAM). Combined with respective advantages of two modeling, a hybrid modeling algorithm is presented from the perspective of multi-scale. The infrared image is decomposed into multi-scaled sub-images by HAM algo-rithm. And then the blind-pixel is restored through the nonparametric regression estimation model in intra-scale; meanwhile the image is upsampled by the image parameter model in inter-scale. HAM establishes a set of mul-ti-scale blind-pixel compensation method based on hybrid parametric and nonparametric regression model. The experimental results show that, compared with the classical compensation algorithms, the algorithm has the ad-vantages of good adaptability and can effectively compensate both isolated and blocked blind pixels. And the compensated image achieves noticeable resolution and edge details.%针对红外成像非均匀性产生的盲元与盲元簇问题，提出一种混合自适应回归的红外盲元补偿算法(HAM)。首先对红外图像进行多尺度分解，并对各分解尺度构造非参数回归补偿；然后对各尺度间构建自回归参数模型实现从低分辨到高分辨的学习，进一步提高补偿精确度。 HAM算法打破了现有补偿算法基于滤波和插值的传统思路，建立了基于多尺度分析的混合自适应回归补偿的新方法。实验结果表明，相比于传统的红外盲元补偿算法， HAM算法具有很好的适应性，对于具有大量孤立和盲元簇图像均能取得很好效果，且补偿后图像具有较好的清晰度与边缘细节。