Quantitative Approach on Image Fusion Evaluation

Image registration and fusion are conducted using an automated approach, which applies automatic adaptation from frame to frame with the threshold parameters. Rather than qualitative approach, quantitative measures have been proposed to evaluate outcomes of image fusion. Concepts of the discrete entropy, discrete energy, mutual information and information redundancy have been introduced. Both Canny Edge Detector and control point identification are employed to extract retinal vasculature using the adaptive exploratory algorithms. The shape similarity criteria have been used for control point matching. The Mutual-Pixel-Count maximization based optimal procedure has also been developed to adjust the control points at the sub-pixel level. Then the global maxima equivalent result has been derived by calculating Mutual-Pixel-Count local maxima. Both cases of image fusion practices are satisfactory whose testing results are evaluated on a basis of information theories.