Research On Image Fusion Method Of Panchromatic Image And Multi-spectral Image Based Wavelet Transform

The aim of image fusion is to combine different images from different sensors and gel a more correct, clear and complete knowing about the target than a single sensor. Application of multi-sensor image fusion span a broad range that includes military and civilian. Multi-sensor image fusion is one popular research field in image processing.This thesis studies the fusion of Panchromatic image and Multi-spectral image based Wavelet Transform. Currently available image fusion methods are not efficient for fusing different multi-spectral image and panchromatic image. To resolve this problem, a new self-adaptive method for fusing different images is introduced in this paper Each image was decomposed into certain sub-images, which are consisted of an approximate image and a set of spatially-oriented detail images by the tow dimension orthogonal wavelets. These approximate image and detail images were merged with different fusion model, which could keep the spectral feature of the original multi-spectral image and improve the content of information effectively, moreover, it has stronger self-adaptive activity.We also analyze and summarize the relation of fusion result and fusion parameter according to valuation methods proposed in the paper. Fusion parameter includes wavelet radix, the number of decomposition levels, different fusion rules, the size of region for measure, different IMS transform methods. At last we get reliabie experimental results and provide useful basis for others fusion methods based Wavelet Transform.Another main contribution of the thesis is the image registration which is an important step for image fusion. Currently available image registration methods are not efficient for large rotation angle image. To resolve this problem, we developed a new self-adaptive registration method with the estimation of rotation angle, which is an accurate and automated image registration methods for multi-spectral image and panchromatic image, especially this method can get a better result for large rotationangle image. First, the wavelet analysis technique was used to extract the corner of two images. Second, the maximum of cross-correlation principle with the estimation of rotation angle is involved to adaptively determine the optimal registration parameters. Finally, we get high accurate registration result by employing relax-tense iterative algorithm which can improve registration precision.