Using Discrete Wavelet Transform to increase the Accuracy of Hyper Spectral and High Resolution Images Fusion

Abstract

In optical remote sensing, hyper-spectral (HS) image which contains color information is produced by<br />hundreds of spectral bands. Because of the trade-off imposed by the physical constraint between spatial<br />and spectral resolutions, the HS image has poor spatial resolution. On the contrary, the panchromatic<br />(PAN) images have high spatial resolution but no color information. Image fusion can combine the<br />geometric detail of PAN image and the color information of the HS image to produce a high-resolution<br />HS image. The aim of this study is the fusion of Hyperion and OrbView-3 PAN images based on Discrete<br />Wavelet Transform (DWT). Firstly, the preprocessing methods were applied on Hyperion and OrbView-3<br />images and registration method based on nearest neighbor method were applied on two dataset. In order to<br />fit images pixel size, the resampling operation was applied. PAN image was decomposed by DWT and<br />then fused by hyper spectral image with GST algorithm. DIV, CC, Q and RMSE accuracy assessment<br />methods were used on final fused image to evaluate the results accuracy. The results showed that using<br />DWT based decomposition PAN image, preserve the spatial information during fusion rule. Also this<br />technique gains high accuracy in term of spectral information of hyper spectral image.