抑制寄生波纹的红外图像降晰函数辨识算法

提出了一种针对军事目标红外模糊图像复原的降晰函数辨识算法。该算法根据气动光学效应形成湍流流场的机理建立相应的模型,将点扩展函数简化为可用参数描述的高斯函数形式;利用红外图像的边缘梯度变化特性提出边缘清晰度改善量,并作为降晰函数参数辨识的评价标准,清晰度改善量取最大值时对应的参数就是观测图像的最佳降晰函数参数;针对复原过程中可能出现的振铃效应,运用细节规整化思想衍生的加权空间复原算法,自适应地抑制寄生波纹的产生。实验验证表明,本文方法能有效地复原红外模糊图像,且对降晰函数的辨识准确率高,相对误差可以降低至4.5%左右。另外,抑制振铃寄生波纹效果良好。复原后图像在各项质量评价指标上都有很大提高,峰值信噪比提高量超过9.4dB,综合评价指数ImageQ提高了20以上。

Blur identification algorithm to suppress parasitic ripples on infrared images

A blur identification algorithm for infrared fuzzy image restoration in military targets was proposed. An appropriate model was established based on the formation mechanism of an aero-optical turbulent field. The point spread function was simplified to a Gaussian function form that could be described by parameters. On the basis of infrared images' edge gradient characteristics, the edge definition improvement was proposed to estimate the blurring parameters and to be acted as the standard of blur identification . When definition improvement was taken its maximum, the corresponding blurring parameter was the optimum parameter of an observation image. For the ringing effects occurred in restoration processing, weighted space restoration algorithm derived from detail regularization was used to adaptively suppress the parasitic ripple. Experimental results show that the algorithm effectively recovers infrared blur images, accurately identifyies the blurring function and the relative errors have reduced to 4.5%.Moreover, the algorithm suppresses perfectly ringing effects during the image restoration process. Data show that all the quality evaluation indicators have been improved, the Peak Signal to Noise Ratio has improved by 9.4 dB and the comprehensive evaluation index ImgeQ improved more than 20.