| The problem on the fusion of Multi-Spectral images and its applications is discussed in this dissertation. Firstly, it shows detail the definition, origin, development and statues of the image fusion. Then, the concept and meaning of Multi-Spectral image is put forward. The aims and characteristics of the Multi-Spectral image fusion are also summarized in the dissertation. The methods of pixel-level image fusion, feature-level image fusion, decision-level image fusion and classification is discussed in this dissertation.Pixel-level image fusion is the basis on Multi-Spectral image fusion. The object of the fusion is low spatial resolution Multi-Spectral image to high spatial resolution gray image. Shah average method, Brovey image fusion method, and high pass filter image fusion method are discussed in this dissertation.Feature-level Multi-Spectral image fusion need to consider the characteristic of spectral. An improved image fusion method based on HIS transform is proposed. In the process of fusion, a multiwavelet transform are embedded. It can make the fused image with high spatial resolution. At the mean time, it can reserve the original spectral. Otherwise, another feature-level image fusion method- Principal Component Analysis is also discussed. It uses K-L transform to achieve the fusion of many low spatial resolution multispectral images with high spatial resolution gray image.In the last of this dissertation, the classification based on decision-level image fusion is discussed. Composite artificial neural network to achieve classification is used. It can be proved that this method has not only high ratio of correct, but also good characteristic of antigambling.More over, the evaluation method of the effect on Multi-Spectral image fusion is introduced detailedly. The applying principal of subjective methods and objective methods are introduced, respectively. The objective evaluation methods are summarized and analyzed especially, and the method of choosing them is proposed. |
PDF Full Text Request