Title: 小波轉換應用於SPOT影像融合之分析
Applying Wavelet Transform for the SPOT Image Fusion
Authors: 董倩琪
Chiann-Cyi Dong
Dr. Tian-Yuan Shih
Keywords: 小波轉換;影像融合;Wavelet Transform;Image Fusion
Issue Date: 1998
Abstract: 影像融合的目標在結合高空間解析度與高光譜解析度之影像,以產生高解析度之多光譜影像,並使其保存原全色態影像之空間資訊與原多光譜影像之光譜特徵;融合後的影像含有良好的地形細節與光譜資訊,能幫助特徵物的擷取與影像之判讀。由於SPOT遙測衛星影像在解析度與幾何條件的優勢,以及對多時影像與立體像對的提供,使得其在影像判讀之研究上大為有利。故本研究將以SPOT影像為試驗對象,進行影像融合方法的分析研究。 小波轉換在影像處理中,具有多層解析度、正交、快速運算及良好定位等特性,適合用來偵測影像中之高頻細節。故本研究將使用二維離散小波轉換與多解析度小波分解之技術,融合SPOT之全色態影像(10m解析度)與多光譜影像(20m解析度),提出一產生10m解析度之多光譜影像的方法。並就空間資訊及光譜特徵上與HPF法及IHS法相比較。 研究結果顯示,以小波轉換進行影像融合,比其他方法更能保存原多光譜影像的光譜特徵,且融合影像亦增加了原全色態影像中之空間資訊。使用小波轉換進行影像融合,所保留的光譜特徵與空間資訊量,可隨著解析度階層之選擇而作調節;因此,以小波轉換進行影像融合,在光譜及空間資訊之保存與取得上有較大的彈性與潛力。
In order to preserve both the spatial information of panchromatic images and the spectral characteristics of multispectral images, high spatial resolution image is fused with high spectral resolution image to generate high resolution multispectral image. The fused image which provides good terrain detail and useful spectral information, facilitates feature extraction and image interpretation. The use of SPOT data for various interpretive purpose is facilitated by the sensing system with excellent spatial resolution, geometric fidelity, and the provision for multidate and stereo imaging. This thesis compares different methods used for the fusion of SPOT XS and PAN images. Wavelet transform has the properties of multiresolution, orthogonality, quick calculation and good localization capability. This thesis investigates the applications of two-dimensional wavelet transform and multiresolution wavelet decomposition technique for image fusion. Methods based on wavelet transform are compared with HPF and IHS methods in terms of both spatial and spectral quality. Compared to the HPF and IHS methods, wavelet transform-based methods preserve the spectral characteristics of multispetral image better and also introduce the spatial information from panchromatic image. With the wavelet fusion technique, it is easy to control the trade-off between the spectral information and spatial structure by adjusting the wavelet decomposition level. Experiments indicate that the wavelet-based methods provide greater flexibility and higher potential for preserving spectral and spatial information.
Appears in Collections:Thesis