摘要
针对目前多聚焦图像融合中融合图像边缘整体性较差的情形,为更好地表示边缘大尺度轮廓和小尺度细节信息,依据离散不可分剪切波能更好地捕捉高维空间奇异性特征,避免频域重叠,而且可以得到更好的框架界和有更好的方向敏感性等优点,利用其对原图像进行分解以保留更多图像细节特征信息;同时利用引导滤波对边缘的平滑能力,并减弱过平滑影响,提出迭代的自适应引导滤波融合规则,通过加权引导滤波的参数让其自适应图像的平坦区域和边缘区域。对于分解后的系数采用基于边缘算子和图像对比度信息的迭代自适应引导滤波优化,较好地保留边缘整体轮廓和小尺度细节特征,降低光晕效应的影响,同时保证边缘连续性较高。最后通过逆离散不可分剪切波变换得到融合后的图像。通过在不同数据集上的实验结果表明,该算法能够更好地保留图像边缘的整体强度和小尺度细节特征,取得了较好的融合效果。
Aiming at the poor integrity of fused image edges in multi-focus image fusion at present,in order to better represent large-scale contours and small-scale details of edges,according to the advantages of discrete non-separable shearlet transform,such as better capturing local features in high-dimensional space,avoiding overlap in frequency domain,better frame boundaries and better direction sensitivity were taken to decompose the original image to preserve more image details and feature information.At the same time,using the ability of guiding filter to smooth edges and weaken the influence of over-smoothing,an iterative adaptive guiding filter fusion rule was proposed,which adapted image flat area and edge area by weighted guiding filter parameters.The iterative adaptive guided filter based on edge operator and image contrast information was used to optimize the decomposed coefficients,preserving the overall contour and small-scale details of the edges,and reducing the halo effect and ensure the high continuity of the edges.Finally,the fused image was obtained by inverse discrete non-separable shearlet transform.The experimental results on different data sets show that the algorithm can better preserve the overall strength and small-scale details of image edges,and achieve better fusion results.
作者
舒衡
古乐野
陈蔓
SHU Heng;GU Leye;CHEN Man(Chengdu Institute of Computer Application,Chinese Academy of Sciences,Chengdu Sichuan 610041,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《计算机应用》
CSCD
北大核心
2019年第S02期236-241,共6页
journal of Computer Applications
基金
四川省科技厅科技成果转化项目(2014CC0043)
四川省科技创新苗子工程项目(SCMZ2006012)
关键词
多聚焦图像融合
离散不可分剪切波变换
迭代自适应引导滤波
边缘算子
对比度
multi-focus image fusion
Discrete Non-separable Shearlet Transform(DNST)
iterative adaptive guided filtering
edge operator
contrast ratio
