Research On Sar, Infrared And Visible Images Registration And Fusion Algorithms

The technology of multisource remote sensing image fusion can fully take the advantages of kinds of remote sensing information acquired by different sensors, and provide valid solutions against the shortage coming from the extraction and interpreting of the single source remote sensing information. The key issue we need to solve before image fusion is image registration.In this paper we go into the research of SAR, infrared and visible image registration under different image resolution, and discuss the multiscale and multidirection wavelet edge interactive variance algorithm based on the region of interest. Groups of simulate results based on the real image verify the effectiveness of the registration algorithm, and then we give the image registration evaluation criterias. For the same or different type of image registration the proposed algorithm has a higher universality and meanwhile it has a better result against the image noise.In addition, we discuss the multisensor image fusion technology, we propose an algorithm based on the SWT and Parcel region label, and an algorithm based on DT-CWT and multilevel Parcel region label. The two algorithms combin the pixel level image fusion with the feature level image fusion, extract the information such as edge and region in each image and generate fusion decision map by fusion strategy to guide the wavelet coefficients fusion. Groups of simulate results based on the real image verify the effectiveness of the proposed algorithms as well. Compared with the classic fusion algorithms based on wavelet transform, the algorithms in this paper have bigger superiorities in edge localization, region consistency and fused image contrast.