基于改进视觉显著性算法的运动模糊图像复原  

Motion Blur Image Recovery Based on an Improved Visual Saliency Algorithm

作　　者：严家奕 侯幸林 孙涛 经正 徐百川 刘德志 YAN Jiayi;HOU Xinglin;SUN Tao;JING Zheng;XU Baichuan;LIU Dezhi(School of Automotive Engineering,Changzhou Institute of Technology,Changzhou 213032;ENN Energy Logistics Co.,Ltd.,Langfang 065000)

机构地区：[1]常州工学院汽车工程学院,江苏常州213032 [2]新奥能源物流有限公司,河北廓坊065000

出　　处：《常州工学院学报》2023年第6期20-29,共10页Journal of Changzhou Institute of Technology

基　　金：国家自然科学基金项目(62101074)。

摘　　要：针对模糊尺度较小时运动模糊参数估计算法对点扩散函数求解不准确的问题,提出基于最小二乘直线拟合的精确角度估计算法,并对微分自相关长度估算方法进行改进,获取了高精度的运动模糊参数。基于视觉显著性加权算法,设计一种无参考图像质量评估系统来确定图像模糊程度以及复原图像的最优解。实验结果表明,所提方法能更准确地求解运动模糊参数,使用求解结果恢复的清晰图像在主观和客观方面优于现有方法,具有较高的工程应用价值。For the problem of inaccurate solution to the point diffusion function,the exact angle estimation algorithm based on least squares line fitting is proposed in small fuzzy scale to improve the differential autocorrelation length estimation method and obtain the high precision relative motion length.Based on the visual saliency weighting algorithm,a reference-free image quality evaluation system is designed to determine the degree of image blur and the optimal solution to the image recovery.The experimental results show that the proposed method can solve the motion blur parameters more accurately,and the clear images reco-vered by the solution results are better than the existing methods in subjective and objective aspects,which has high engineering application value.

关 键 词：运动模糊图像复原 运动模糊参数估计 最小二乘直线拟合 视觉显著性加权 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]