The Research On The Pixel Level Image Fusion Technique And The Analysis Of Its Performances

Image fusion is to get a more accurate, comprehensive and reliable image description of the same target through using the information of multiple images and by certain algorithm. It plays a significant role in remote sensing, medicine, natural resource exploration, oceanological management, biology and etc. and it also has strategic significance in national defense security and economic construction. Therefore, its further research has theoretical significance and wide prospect of application in military and non-military fields.Image fusion consists of such three levels as the pixel level, the feature level and the decision level, among which the pixel level image fusion can provide more abundant, accurate and reliable detailed information that doesn't exist on the other levels and It is the most complicated in the whole image fusion techniques and also is the most difficult to implement in the fusion processing techniques this dissertation progresses mainly around the pixel level image fusion and proposes a variety of pixel level image fusion techniques according to the key problems in the pixel level image fusion techniques. The major research and findings are as follows:(1)This dissertation deduces and generalizes the algorithms of the existing mainstream pixel level image fusion in details and realizes the Matlab experiment and adopts the contrast analysis to analyze the research findings. Then it analyzes the performance of each algorithm of the four kinds of images, that is the multi-focus images, the remote-sensing images, the infrared images, the front-shot infrared images and the dim-light night-viewing images, and gets the optimum algorithm corresponding to each existing algorithm; secondly,(2) This dissertation probes into the image fusion quality assessment and deduces a set of indexes as the criteria to analyze the performances of this discussion.(3) It further studies the image fusion technique based on the wavelet transform and through mass experiments on three different kinds of images, this dissertation finds out the corresponding optimum decomposing layer to each other and optimizes the fusion effect.(4) Based on the region entropy maximization, this dissertation puts forward a kind of improved image fusion technique, which adopts image region entropy as fusion operator, and adopts the matching degree of the region entropy as the fusion threshold .Compared with the image fusion algorithm based on the pixel principles, the experiment proves that the image fusion algorithm based on the region entropy maximization in this dissertation produces better fusion effect and has certain practicability.