Image Fusion And Super-resolution Reconstruction In Multiscale Transform Domain

Image plays an important role in the process of human acquiring and transmitting information.Single type of imaging sensor cannot meet the needs in practice for the limitation of imaging sensors.Therefore,the problem is how to acquire a more real image with abundant features so as to meet the needs of related fields.The fusion algorithm can acquire a image having rich details and more real through fusing two images or more with complementary information.In addition,some areas require high resolution images such as medical,military and so on,and the quality of images acquired by imaging sensors is not satisfied.So the problem is how to get higher resolution images when the imaging sensor is unchanged.The image's super-resolution reconstruction algorithm can restore a blurred image into a rich and informative image.Under the above circumstance,the dissertation concentrates on the study of fusion algorithm and single image super-resolution.The main contents of this thesis are as follows:1.To overcome the shortcomings of traditional remote-sensing image fusion methods,a novel image fusion method based on Non-Subsampled Dual-Tree complex Shearlet Transform(NSDTST)is proposed in this paper.In our fusion framework,the NSDTST is firstly performed on each of the source images to obtain their low-pass and high-pass coefficients.Then,the low-pass bands are merged with a Multi-Feature-Self-Similarity(MFSS)fusion approach;For the high frequency sub-band coefficients according to its directional feature,a improved fusion method based on adaptive directional Pulse Coupled Neural Network is presented.The experimental results show that the proposed algorithm is better than others to balance the spatial resolution and spectral resolution.2.For the traditional super-resolution algorithms based on Sparse Coding(SC),a novel reconstruction algorithm is proposed which takes the advantages of Nonsubsampled Pyramid Transform and Pulse Coupled Neural Network(PCNN).Firstly,the high and low frequency subband coefficients are obtained by NSPT.For the low frequency sub-band,we use the improved PCNN to get the best coefficient of SC algorithm.For the high frequency sub-band,we reconstruct it by using the improved PCNN.The experiment shows that the algorithm has good results in both subjective and objective data.