基于动态卷积模板的非均匀光线图像复原算法研究  

Imbalanced Illumination Image Restoration Based on Dynamic Convolution Mask

作　　者：刘星辰 朱晋 LIU Xingchen;ZHU Jin(Jilin Meteorological Information Network Center,Changchun 130062;Yanbian University,Yanji 133002)

机构地区：[1]吉林省气象信息网络中心,长春130062 [2]延边大学,延吉133002

出　　处：《现代计算机》2021年第18期113-116,共4页Modern Computer

摘　　要：针对不均匀光线图像的复原问题,本文提出了一种非静止的卷积模板构造算法,利用像素点与其邻域的强相关性为每一个像素点构造不同的卷积模板。使用这种动态模板,在完全未知相关原始图像和图像退化模型的情况下,对退化图像进行盲图像复原;本文还提出了一种曝光过度退化模型,用于模拟真实情况中的曝光过度现象。实验结果表明,本文提出的图像复原算法能够修复由于光学成像缺陷造成的曝光过度的图像,改善图像细节,提高图像质量;经本文提出曝光过度退化模型模拟的曝光过度图像效果良好。To solve the problem of imbalanced illumination image restoration,this paper proposes a nonstationary convolution mask construction algo⁃rithm.Using neighborhood information,a changing convolution mask for each pixel is calculated to restore the image.Using this nonstation⁃ary mask,blind image restoration is performed without any information about original image and image degradation model.A light saturated regions degradation model is proposed to simulate the overexposure phenomenon in real world.The experimental results show that the pro⁃posed image restoration algorithm can restore the over-exposure image caused by optical imaging defects,improve image details and quali⁃ty;the effect over-exposure image simulated by the light saturated regions degradation model proposed in this paper is good.

关 键 词：图像复原 图像滤波 盲卷积 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]