Research Of Multi-source Remote Sensing Information Fusion

In recent years, with the development of remote sensing technology, computer technology and information processing technology, multi-source remote sensing information fusion technology is utilized successfully in a broad variety of fields including military, remote sensing, automatic target recognition, machine vision, medical image processing and etc. Multi-source remote sensing information fusion is defined a new technique to match two or more images of the same scene taken at different view points or by different sensors, and then integrate the images as one image in order to obtain fusion result which has more reliability, less ambiguity and better understanding.The existed algorithms, such as IHS transform fusion, PCA transform fusion, HPF transform fusion, pyramid fusion, wavelet transform fusion and the combined fusion algorithm due to their own characteristics and complexity, usually are with the problems which are not ideal in their clarity and reservation to the original information. In this research, for the problems mentioned above, an improved image histogram based on Gaussian function matching algorithm is proposed. It makes the image gray value has a very good balance and stretch to achieve good matching effects. Then, based on the traditional wavelet packet fusion algorithm, an improved multi-channel filter wavelet packet fusion algorithm is proposed. Before high spatial resolution images decomposed, multi-channel filtering retained the good scale features, and it makes an improved the adaptive fusion criteria based on regional characteristics. Information on high-frequency and low frequency uses different fusion criteria, improving the spatial resolution better. Finally, in the use of world-renowned remote sensing software ENVI 4.5 and ERDAS IMAGINE 9.2, the existed and improved fusion algorithm are compared in experiment of this research. From both the visual subjective and statistical objective evaluation, the fusion results obtained carried out some analysis and summaries. Compared with the existed fusion algorithms, the fusion result of the improved fusion algorithm of this research not only enriches the amount of information, retains a large number of texture features and spectral characteristics, and is with color fidelity, better integration and higher-definition.