الفهرس | Only 14 pages are availabe for public view |
Abstract Medical image processing has captured a large interest from the researchers in the last decade. With the development of different medical imaging modalities, it has become a must to find new image processing tools and algorithms that can help in the automated medical image diagnosis task. Of such tools, medical image fusion finds high popularity. Image fusion, in general, is the process of integrating information existing in two or more images for the same scene in a single image. This concept has been developed to some extent to include the fusion of multi-modality images in medical applications. It is known that different medical imaging techniques work with different principles, and hence they give images with different characteristics. That is why the medical image fusion task is a complex task that needs image registration, image dynamic range modification, and efficient fusion algorithms to integrate the obtained images in the fusion results. This thesis is concerned with this complicated task of medical image fusion. It covers the main concepts and algorithms utilized form medical image fusion. In addition the quality metrics that can be used for the assessment of medical image fusion results are also covered. The thesis presents two proposed techniques to achieve better medical image fusion. The first technique is a hybrid technique that merges wavelet-based image fusion with principle component analysis image fusion. An adaptive strategy is adopted to switch between both techniques according to the image local activity levels estimated with the local variance. This proposed technique achieves better quality indices than those obtained with the traditional techniques. The second proposed technique depends on the utilization of optimization concepts in order to maximize the image quality metrics after the fusion process. This technique comprises histogram processing, image registration, and wavelet-based image fusion. It is carried out iteratively towards the best solution from the quality metrics perspective. Simulation results prove the best quality of the fusion process with the proposed technique that depends on optimization with an acceptable processing time taken through the optimization process. Finally, both suggested techniques can be used in a unified framework for automated medical diagnosis that can help specialists to take the correct decision according to the obtained high-quality fusion results. |