Research On Image Fusion Algorithms Based On Wavelet Transform And Region Segmentation

Multi-source Image Fusion is an important and useful subject in Image Understanding and Computer Vision. In recent decades, image fusion method in wavelet domain is becoming a most active and challenging field, which is the focus of this thesis. The main contributions are as follows:1. Some traditional image fusion methods based on 2-band wavelet transform and M-band wavelet transform are studied and simulated. Some problems of these methods are analyzed.2. Region-based image fusion using clustering-based image segmentation is studied. A new region activity level criteria and a new decision-map filter are presented. Fusion experiments using this activity level criteria and decision-map filter in 2-band wavelet domain and M-band wavelet domain are carried out. The analyzed results show this scheme can improve many evaluation criteria and visual quality and is robust to misregistration.3. An efficient graph-based image segmentation algorithm is introduced into the region-based image fusion. A new fusion scheme are presented and simulated. The analyzed results show this new fusion scheme can improve many evaluation criteria and visual quality and has superiority over window-based fusion scheme and cross-band-window fusion scheme.4. Multi-source Image Fusion System based on wavelet transform and region segmentation is designed and implemented. Class libraries of wavelet transform, region segmentation algorithms, image fusion algorithms and objective evaluation class are provided.5. The proposed image fusion algorithms are summarized and compared. The prospect of image fusion in wavelet domain is analyzed.