Research On Image Fusion Based On Multi-Scale Decomposition

As one field of information fusion, image fusion has been extensively applied in multi-focus images, remote sensing images and space images. This paper focuses on the methods of image fusion based on multi-scale decomposition. The main contributions can be summarized as the follows:(1) The pretreatment of image fusion is accomplished in the thesis. The image denoising methods and the threshold selection methods of wavelet denoising are studied. An improved denoising method is proposed based on the wavelet denoising. This method not only keeps the continuity of the denoising function and the advantages of the soft threshold denoising method, but also keeps the less image reduced characteristics of the hard threshold denoising method. The denoising effect is analyzed by the experiments.(2) The method and the steps of wavelet transform image fusion are provided. On the basis of shift invariant discrete wavelet transform, a new rule of the adaptive weighting coefficients based on region energy difference is proposed in the high-frequency sub-bands. A combination of region energy maximum method and the weighting coefficients method is proposed in the paper. The adaptive weighting coefficients can be determined by region energy difference. The relationship between the threshold T and image evaluation is analyzed and the improved effectiveness of the new fusion rule is shown by the experimental results.(3) The nonsubsampled contourlet transform (NSCT) is studied in the paper. The steps of image fusion based on countourlet and NSCT are provided. An improved image fusion rule of low-frequency and high-frequency sub-bands based on NSCT is proposed. On the basis of NSCT, the maximum method of standard deviation is applied in the low-frequency sub-band, and the region contour energy contrast method is applied in the high-frequency sub-bands. The final decision of the high-frequency sub-bands are the joint decision of the low-frequency sub-band and high-frequency sub-bands. The improved effectiveness is analyzed by the experiments.(4) The image fusion tools are designed and implemented. Pretreatment module, image fusion module and image evaluation module are developed and image fusion is realized.