基于超分辨率融合的双路径亚像元定位

A dual path subpixel mapping method based on super resolution fusion

摘要:

为了达到精确高光谱图像(hyperspectral images,HSIs)亚像元定位(subpixel mapping,SPM)结果,本文提出基于超分辨率融合(super-resolution fusion,SRF)的双路径亚像元定位方法(super resolution fusion dual path subpixel mapping,SRF-DPSPM)。在该方法中采用两种路径处理高光谱图像。在一条路径中,对原始高光谱图像的丰度图像进行插值操作,得到高分辨率的丰度图像；在另一条路径中,采用SRF技术,利用多光谱图像(multispectrum images,MSIs)对原始图像进行处理,得到新型高分辨率的丰度图像,融合两种路径下的结果。由于结合了插值路径和超分辨率路径下的丰度图像,从而得到具有丰富空间-光谱信息的预测值。最后,根据预测值得到最终的亚像元定位结果。在两个真实数据集上进行实验分析,结果表明,该方法能进一步提高精度。

Abstract:

In order to achieve accurate subpixel mapping (SPM) results,this paper proposes a dual path subpixel mapping method for hyperspectral images (HSIs) using super-resolution fusion (SRF) technology.In this method,two paths are used to process hyperspectral images.In one path,the abundance image of the original hyperspectral image is interpolated to obtain a high resolution abundance image.In the other path,the super-resolution fusion technology is used to process the original image using multispectral images (MSIs) to obtain a new high resolution abundance image,and the results of the two paths are fused.Because the abundance images under the interpolation path and the super-resolution path are combined,the predicted values with rich spatial-spectral information are obtained.Finally,the final sub-pixel mapping results are obtained according to the predicted values.Experimental analysis on two real datasets shows that this method can further improve the accuracy.