Fusion Technology Based On Multi-Spectral Image And Panchromatic Image

The so-called images fusion integrates images information, which is collected by two or more different sensors at the same or different time in accordance with certain algorithmic rules. In that way, we can obtain flusher, more reliable and more accurate image information comparing with the single image. The present research will facilitate further analysis and understanding of the image fusion, which is widely used in military and civilian fields.This thesis first introduced the commonly used image evaluation, the traditional image fusion methods and then analyzed the advantages and disadvantages of each algorithm. The composite fusion algorithm is proposed based on their respective merits and demerits. Principal component analysis and image fusion methods that enhance the wavelet are put forward after evaluating the shortcomings of wavelet image fusion. Finally, the basic rationale of Contourlet transformation is presented by proposing the combination of principal component analysis and Contourlet transforming image. Contoulet transformation cannot satisfy the translation invariance, resulting in a fusion of images produced Gibbs effect. In order to overcome this defect, this thesis introduces the rationale of non-subsampled Contourlet transformation and its application in image fusion.Composite fusion method is put forward to effectively combine low-resolution multispectral image and high-resolution panchromatic images. After the decomposition of wavelets, low-frequency component selects the highest absolute value and high-frequency component the biggest regional variance and also compares with the classical fusion algorithm. Afterwards, this thesis poses the combining technique between principal component analysis transformation and the upgrading of wavelet, which is closely followed with the fusion of high and low frequency components regarding the wavelet lifting based on different fusion rules. Moreover, comparison between PCA+DWT and HSV+LWT fusion methods are also conducted in this study. The experiment indicates the abovementioned methods both not only reserve multispectral image as much as possible but also improve the spatial resolution of multispectral images.The present thesis proposed image fusion method that combines the principal component analysis and the Contourlet transformation in the end. After the analysis and transformation of the multispectral image, we transformed the first principal component and the full-color images. A new first principal component was gained by using the low-frequency components with the weighting method based on fuzzy membership and high frequency components with the energy threshold matching method. Finally, the new first principal components and other principal components are combined to give the final fused image, Gibbs effect in order to eliminate the fused image. Furthermore, the study introduced the combination fusion method that contains Non-subsampled Contourlet transformation and HIS, which fuses the low-frequency components by using principal component analysis and effectively overcomes the deficiencies of the traditional methods in the low-frequency components. The experiment shows that this method not only retains the spectral characteristics of the multispectral images but also improves the spatial resolution.