SWT Image Fusion Based On Phase Congruency Image Registration

Synthetic Aperture Radar(SAR) which provides high-resolution images in all-weather, all day and all night is widely used in military reconnaissance and civil field. SAR image registration and fusion is an important image processing technology in remote sense field,many researchers pay close attention to it in recent years. Image registration is the process of matching two or more images that get from the same scene derived from different time,different sensors or different views of angle. And image fusion technique is used to reconstruct a new image, which provides richer visual information than the original images,with the help of complementary data information of multi-sensors. In fact, image registration is a key step and the necessary prerequisite of image fusion.SIFT(Scale Invariant Feature Transform) operator is scale invariant,rotation invariant,Thus, It has been successfully widely used in the field of image registration. But, due to the complexity of SAR imaging system, variety of SAR image's characteristics, the typical SIFT algorithm based on the information of local neighborhood gradient, The SIFT algorithm based on the information of local neighborhood gradient,when against the SAR images which are poor quality or large difference in gray and contrasts, detect feature directly by the difference of the pixels' gray level, which is notstable enough, leading to a low precision or failure. Considering the above problems, this paper firstly preproccesses original SAR image by Lee filtering to suppress speckle noise making use of statistical properties of SAR image speckle. Then this paper takes advantage of the phase consistency model to extract image feature point, in this way it can reduce impact of unstable and wrong feature points in SAR image. Later a new feature point matching strategy is proposed. The strategy can ensure operational speed and greatly eliminate false match points. Then this paper uses ransac algorithm to obtain affine transformation model of an image; Finally, a bilinear interpolation method is used for image resampling. With the real SAR image simulation results, it shows the proposed method has high registration accuracy,fast speed,and strong robust properties.As a result, it can effectively complete SAR image registration tasks.In addition, this paper introduces the basic theory and the commonly used method of image fusion technologies. Due to the translation invariance of discrete Wavelet Transform in image fusion, the method of Stationary Wavelet Transform(Stationary Wavelet Transform,SWT) to realize the image fusion is proposed.This article first advocates SWT decomposition of image, then for target contour, texture, color information and so on the performance of different characteristics of SAR and optical images, then puts forward a strengthen the complementary information fusion rule, that is low frequency part of image is based on weighted rule of edge detection and high frequency part adopts absolute value bigger first principle,to implement information fusion between images of high frequency and low frequency. Finally using SWT inverse transformation to get image fusion results.With simulation fusion experiment of SAR, millimeter wave and visible light image, the method of the paper is verified to possesses the advantages of translational invariance, and information and clarity of fused image is improved.