基于多尺度边缘表示的偏振图像二次融合算法

摘要

针对传统成像技术探测信息量少以及传统图像融合方法计算量大,不能凸显边缘轮廓的问题,提出一种基于多尺度边缘表示的偏振图像二次融合方法.首先,对同时偏振成像系统获取的4个偏振角度图像进行多尺度边缘去噪、配准预处理.然后一次融合,即将预处理后的4个偏振角度图像进行二进小波分解,在不同尺度上对低频、高频系数进行特定组合,计算得到Stokes向量[IQUV]T,线偏振度(DOLP)和偏振相位角(AOP)等偏振信息;二次融合,即是根据边缘相关性原则融合偏振强度图像和线偏振度图像经小波分解后的高频系数,低频用区域能量的方法.最后,采用交替投影迭代法重构融合图像的低频、高频系数,得到重构图像.实验结果表明,本文算法的融合性能优于传统方法.以小波变换法为参考,平均值提高65.48％,标准差提高95.98％,平均梯度是传统小波的6.45倍,边缘强度是小波变换的5.54倍.本算法提高了运算速度,结合了强度图像亮度信息和线偏振度图像反映不同物体性质的特点,突出了目标轮廓细节,能很好地识别隐藏、伪装的目标.%The traditional imaging techniques have less information and the traditional image fusion methods need large calculation capability.That result in the weak capability to highlight the edge profile.To solve this problem,a kind of polarization image secondary fusion method based on the multi-scale edge representation was proposed.Firstly,the preprocessings of multi-scale edge denoising and registration were carried out for the 4 polarization images which were obtained from the same polarization imaging system.Then the binary wavelet decomposition was done for 4 polarization images after the preprocessing,and specific combination was done for low-frequency coefficient and high-frequency coefficient at different scales,and then the polarization informations were calculated and gotten,such as Stokes vector[I Q U V] T,linear polarization degree (DOLP) and polarization phase Angle (AOP);the high-frequency coefficients were fused after wavelet decomposition of polarization intensity image and linear polarization degree image based on the edge correlation principle,and the fusion of low-frequency coefficients use regional energy method.Finally,the alternating projection iterative method was used to reconstruct low-frequency and high-frequency coefficient of fusion image,and reconstructed image was obtained.Experimental results show that the fusion performance of the proposed algorithm is better than traditional methods.Taking wavelet transform method as a reference,the average value increases by 65.48％,standard deviation increases by 95.98％,the average gradient is 6.45 times of the traditional wavelet,edge strength is 5.54 times of the wavelet transform.This algorithm improves the calculation speed,has both the brightness information of intensity image and the feature of linear polarization image,which can reflect properties of different objects,highlight the target contour details and can identify hidden,camouflaged targets.