Research On Remote Sensing Image Fusion Algorithm

With the rapid development of the remote sensing technology, multi-source remote sensing image fusion technology has been concerned and applied widely. Its theories and methods have become an important research realm of the subjects such as intelligent information processing, and remote sensing application, etc. Multi-source remote sensing image fusion technique is to conduct space registration on the same area image data originating from different sensors, adopt certain algorithm to combine information advantage or complementarity in each image data organically, and achieve an image with more abundant information. Multi-sensor image fusion overcomes the limitation and otherness of the single sensor image in geometry, spectrum, and space resolution, etc. Now, multi-sensor image fusion technology is widely applied in some fields, such as medicine, remote sensing, computer vision, and automatic object detection, etc.Firstly, this thesis introduces the current situation and developmental tendency of remote sensing image fusion technology at home and abroad, and introduces the basis theory of remote sensing image fusion. It discusses the pretreatment process of image fusion, such as geometry emendation and image registration. This thesis studies several classical methods, such as intensity-hue-saturation transform, principle component analysis transform and wavelet transform, etc. Secondly, fusion algorithm based on multi-resolution thought adopts different fusion rules, which has important influence on the fusion results. This thesis proposes a fusion method of wavelet packet transform based on local deviation. This method executes effective edge detection by high-frequency image adopting the multi-window local deviation maximum fusion rule, and as much as possible reserves the part of the real image edge detail. It combines the characteristics and details which come from different image effectively. This method has good visual effect. Thirdly, aiming at the limitations of the image fusion method based on traditional wavelet transform with higher algorithmic complexity and more auxiliary memory consuming, this thesis presents a fast image fusion algorithm based on multi-window local deviation. This algorithm uses lifting scheme wavelet packet transform as a multi-resolution analysis tool and simplifies the calculation of some parameters. It significantly reduces complexity of the algorithm and obtains a good fusion results. Finally, in allusion to intensity-hue-saturation transform having spectrum aberrance phenomenon in image fusion, the intensity-hue-saturation and wavelet packet transform remote sensing fusion based on regional feature weighting is proposed. It combines the advantages of the intensity-hue-saturation transform and wavelet packet transform, and adopts feature weighting on histogram-matched new panchromatic component and the intensity component of intensity-hue-saturation transformed multi-spectral image, and then conducts wavelet packet fusion. The result shows that the algorithm not only enhances spatial detail information of the high-resolution image, but also preserves spectral information of the multi-spectral image well.