A Research On The Image Fusion Methods Based On Image Sequences

Image fusion is the current research hotspot in the field of image processing, of which the main principle is fusing several images of a target object so as to increase the precision and reliability of the target image. In the field of image fusion, present studies mainly focus on the fusion methods specific to the images of the same perspective,while pay little attention to those aimed at fusing image sequences. In view of this, a thorough and comprehensive study is conducted in this paper to research on the fusion methods of image sequences. First of all, by analyzing current image fusion technologies and common methods, the author puts forward a method for multi-scale image fusion based on Hidden Markov Model(HMM) and Genetic Algorithm(GA).Besides, with the purpose of improving the accuracy of image sequences fusion as well as eliminating ghost, another hierarchical fusion method is proposed on the basis of in-depth information of the image sequences. The main contents of this study are as follows:First, referring to the domestic and international literature reviews on image fusion,this thesis demonstrates the current studies in this field and analyzes the existing problems.Second, focusing on the relevant technologies, this thesis introduces the theoretical basis of image fusion, including its classification of different levels, all sorts of pixel-level fusion methods, image multi-scale transformation method(such as Wavelet Transform, Ridgelet Transform, Contourlet Transform), and evaluation criterion of image fusion quality.Third, by systematically describing the principles, methods and processing procedures of GA, the paper proposes image fusion method based on GA.Fourth, in order to gain more precise and reliable fusion effect, a multi-scale fusion method based on HMM and GA is proposed. Set on GA fusion method, the principle is to establish HMM between pixels and their parent nodes, and then to work out the fusion coefficients layer by layer with the guide of GA. With the obtained coefficients,the HMM will be renewed, and guide next round fusion.What’s more, a fusion method based on the in-depth information of the images is also put forward to deal with the fusion problems about image sequences. The methodtakes good advantage of the displacement changes of the taken image sequences, and calculates the disparity between the two images and then transforms them into depth map. With the help of these in-depth information, the foreground and background images are separated. According to the different characteristics of the foreground and background images, different fusion rules are applied to fuse the images.