Research On Application Of Shearlet Transform In Image Fusion

Recently, image fusion is widely applied in various fields of machine vision,pattern recognition, remote sensing and so on. In this paper, shearlet transform isapplied to image fusion, and the new algorithms are researched. Experimental resultsshow that both in visual effect and objective evaluation criteria, the presented methodoutperforms traditional methods.Firstly, the algorithm of multi-focus image fusion is researched. According to theimaging characteristics of the multi-focus images, the algorithm of adaptive fusion ruleswhich based on shearlet transform is proposed. In this algorithm, the parameter iscontrolled by both average gradient and space frequency of image. In the process ofmedical images fusion, according to the imaging characteristics of the medical images,the fusion rules of regional matching degree is proposed. It is verified that the proposedalgorithm has better clarity and richer details information.Secondly, the algorithm of the visible images and infrared images fusion isresearched. According to the imaging characteristics of the visible images and infraredimages and human visual system feature, the algorithm of fusion is proposed which isbased on combination of the PCNN and shearlet transform in the fusion process, whichadopt the spatial frequencies as the link intensity of PCNN and combined with imagevisibility. The experiment shows that the proposed algorithm is successful which couldeffective retain the features of source images and be suited to human vision andmachine perception.Lastly, the algorithm of the multi-spectral images and panchromatic images fusionis researched. According to the imaging characteristics of the multi-spectral images and panchromatic images and the characteristics of HSV color space, the images fusionalgorithm is proposed which is based on the combination of shearlet and HSV transform,this algorithm reached to a good balance in the two aspects of spatial resolution andspectral characteristics. The experiments show that the proposed algorithm has betterfusion performance and visual effect.