基于多尺度分析的遥感影像融合研究

Research on remote sensing image fusion based on image multiscale analysis

本文对SPOT5的多光谱波段和全色波段在像素级的融合层次上运用多尺度分析的方法进行了融合试验,主要用了小波变换和Curvelet变换的方法,这两种变换方法都能把图像分解为低频的近似图像和高频的细节图像,采用一定的融合规则对分解后的图像进行融合,并进行反变换得到融合后的图像,并把基于多尺度分析的融合结果与传统的融合方法进行了对比分析。结果表明,基于多尺度分析的融合方法比传统的PCA、Brovey融合方法效果要好;而Curvelet变换融合在光谱保持度及空间信息提高方面都比小波变换融合有所提高。

With the development of the remote sensing, multi-sources images with the same area are more and more rich, we could gain more accurate and richer information from the fused image than from the single image.This paper tests the fusion of multi spectral bands and panchromatic band of SPOT5 image based on the multi scale analysis in the pixel-based fusion level.The main methods are wavelet transformation and curvelet transformation.Both methods can divide the original image into low frequency parts and high frequency parts, then take a proper fusion principle to fuse the decomposed parts, in the end, do a reverse-transformation.To analyze the fusion results, we compare it with the traditional PCA and Brovey method image fusion in the aspects of spectral and spatial quality.The results indicated that the fusion method based on multi scale analysis is better than the traditional and the Curvelet transformation is more suitable for image fusion than the wavelet.