基于可见光的多波段偏振图像融合新算法

Novel Image Fusion Algorithm for Multi-Band Polarimetric Image Based on Visible Light

采用了一种新的基于小波变换的偏振图像融合算法。首先,将两个波段中的每一波段三幅偏振图像利用小波变换分解成低频和高频部分,低频的小波系数平均值作为融合后的低频系数,高频细节系数根据不同区域特征选择方法以及对应输入图像小波系数的窗口区域方差来确定融合后高频小波系数,得到一个波段一幅图像。接着,将得到的图像再进行小波分解,采用低频图像的小波系数最小值作为融合后的低频系数,高频图像根据纹理一致性测度的纹理检测确定融合规则,用来调整高频小波系数,将来自不同图像的特征与细节融合在一起,并对融合图像质量进行了对比评价。实验结果表明,融合后的偏振图像不仅反映了场景的偏振信息,而且还包含了丰富的光谱信息,目标与背景的衬比度也得到了增强,为进一步的目标检测和识别提供了便利。

An novel image fusion algorithm of remote polarimetric image based on wavelet transform is proposed.At first,three polarimtric images in each band of two bands are decomposed into images of low frequency band and high frequency band by wavelet transform.The fused low frequency approximate coefficients are obtained with average method,and the fused high frequency detailed coefficients are obtained with different area-based feature selection method by setting each coefficient equaling the corresponding input image wavelet coefficients that has the greatest multi-window local deviation.One image in 555 band and the other in 665 band are obtained.Then,the algorithm decomposes the two images in one band into parts with low frequency band and high frequency band by wavelet transform,it adopt the minimum low frequency image wavelet coefficients as the fused low frequency image wavelet coefficients,and high frequency image fusion rules based on the texture detection algorithm of texture homogeneity measurement are used to adjust high frequency image wavelet coefficients.The low frequency component and orientation detail component of image are fused pertinently,features and details from different images are fused successfully and the quality of these fusion images is compared.The experimental result shows that the fused images present the polarization information,acquire more spectral information and higher contrast,and object detection and recognition can be more easy with fused images.