多级方向加权最小二乘滤波器及其在多传感器图像融合中的应用

针对多尺度分解在图像融合领域中的广泛应用,本文提出了一种多级方向加权最小二乘滤波器图像多尺度几何分析方法。该方法利用加权最小二乘滤波器对图像进行多级边缘保持分解,得到一个近似图像和多个不同尺度上的细节图像,然后采用小尺寸方向剪切滤波器对细节图像进行方向分析,在不同尺度上生成多个方向细节图像。根据近似图像和方向细节图像所具有的不同物理意义,分别采用不同的融合策略对分解后的图像系数进行合并处理,最后应用多级方向加权最小二乘滤波器的逆变换得到融合图像。多组图像融合实验结果表明,在图像融合领域,本文提出的基于多级方向加权最小二乘滤波器的图像分解方法优于已有文献中的一些典型多尺度分解方法。

Multistage directional weighted least squares filter and its application in multi-sensor image fusion

The multi-scale decomposition(MSD) method is extensively employed in image fusion domain and a novel image multi-scale geometrical analysis(MGA) technique based on multistage directional weighted least squares filter (MDWLSF) is proposed. In the developed method, the weighted least squares filter is utilized on image for multistage edge-preserving decomposition, and an approximate image and some detail images at the different scales are obtained. Then small size directional shear filters are applied to the detail images for the direction analysis, multiple directional detail images are generated at the different scales. On the basis of the different physical meanings embodied in the approximate image and the directional detail images, the different fusion criteria are used to merge the decomposed coefficients respectively. Finally, the fused image can be obtained by using the inverse MDWLSF transform. The fusion experimental results on multi-group different modality images demonstrate that the proposed MDWLSF method is superior to some classical multi-scale decomposition tools introduced in the existing literatures.