| Image fusion is the process that multiple source images of the same target are fusedinto one image. The fused image can obtain more comprehensive and accurateinformation than a single one. With the development of the sensor technology, computertechnology and information processing technology, image fusion has been appliedwidely in the military, automatic identification, machine vision，medical imaging andmany other fields. Shearlet transform is a new multiscale geometric analysis (MGA)tool, it can overcome the shortcoming of Wavelet transform which can not capture thegeometry features of multidimensional data, and Two-dimensional images can be bestapproximation by Shearlet transform.For the imaging characteristics of multi-focus images, remote sensing images,multi-spectral images and panchromatic images, based on the theory of Shearlettransform, three types of image fusion algorithms are proposed in this thesis.The main contributions of this dissertation are summarized as follows:1. According to the imaging characteristics of the multi-focus images, multi-focusimage fusion algorithm based on Shearlet is proposed in this algorithm. Since theShearlet transform can decompose images in any scale and any orientation and it hasastrong ability to express the edge, the Shear transformis introduced into multi-focusimage fusion. Moreover, based on the ratio of sharpness, energy and variance,multi-decision rules for multi-focus image fusion is proposed. This algorithm isbetter to maintain the focus part of the source images, and the whole clear image canbe obtained.2. By the study of pulse-coupled neural network (PCNN), and the imagingcharacteristics of remote sensing images, a fusion algorithm of remote sensingimages based on Shearlet and PCNN is proposed. Due to the fact that Shearlettransform can better capture the features of image details and edges, and the humanvision system to edges is more sensitive than gray value, so the remote sensingimages, combined with Shearlet transform, image gradient information and PCNN,are fused in the proposed algorithm. The algorithm takes full advantage ofmulti-directional characteristics of Shearlet transform and bionic vision system ofPCNN so that the fused image can effectively maintain the goal information ofmulti-source remote sensing images. 3. For solving the distortion problems in the multi-spectral and panchromatic imagefusion, multi-spectral and panchromatic image fusion based on the Shearlettransform is proposed. In this algorithm, based on the idea of regional fusion, andcombining with the concept of the ratio of region sharpness and degree of regionmatch, region-based fusion rule of multi-spectral and panchromatic image isproposed.Images are fused according to this rule. The fused images can not onlymaintain the spectral information, but effectively enhance the spatial resolution. |
