红外图像多尺度统计和应用先验去模糊模型

Deblurring model of infrared image multi-scale statistics and application of prior

为了提高特定应用场景的红外导引头成像质量,采用了统计导引头图像对成像环境和应用场景建模的方法,一方面用L 1/L 2范数对复原图像进行约束,保持多尺度成像细节信息;另一方面用稀疏的拉普拉斯分布对迭代模糊核进行约束,保持对红外成像内容的约束,并采用计算图像细节信息进行了自适应变化核。结果表明,建立的图像复原约束模型能有效地提升成像质量,凸显图像边缘,其对比度增强系数指标提高了20%~50%,峰值信噪比提高了0.8~3.4,图像像素的模糊检测累积概率提高了0.3~0.5。该研究对复杂场景和动载体成像处理有一定的帮助。

In order to improve specific application imaging quality of infrared seeker,a model for imaging condition and application scene was constructed by using statistical image of infrared image seeker.On the one hand,L 1/L 2 norm was used to constrain the restored image according to the characteristics of multi-scale imaging,which kept details in the iterative restoration.On the other hand,a sparse Laplacian distribution was used to constrain fuzzy kernel,and to maintain image’s content.Image kernel size can be adjusted adaptively by calculating the image details.The result shows that the prior constrain algorithm of this paper can effectively improve the image quality.In addition,the evaluation index is improved by this prior design,the contrast enhancement coefficient index is increased by 20%~50%,the peak signal to noise ratio is increased by 0.8~3.4,and the cumulative probability of blur detection is increased by 0.3~0.5.This study is helpful for complex scene and moving vector imaging.