Performance Evaluation of Image Fusion by using Copulas

With the ready availability of multiple sensors, the area of information fusion has been receiving increasing attention. For multi-sensor image data, algorithms such as simple average method, Principal Component Analysis (PCA) method, Gradient Pyramid (GP) method, Laplacian Pyramid (LP), Ratio Pyramid (RP) method and Discrete Wavelet Transform (DWT) methods have been successfully applied for image fusion. Another important issue that arises in image fusion: the performance of image fusion is that the performance of the associated algorithms is difficult to evaluate, especially when a clearly defined ground-truth image is not available. Some common measures to assess the performance for image fusion are Mutual information (MI), Tsallis and Renyi divergence based information. However they are difficult to estimate precisely. In this paper, a new approach is proposed for evaluating the performance of image fusion algorithms based on copula functions. To achieve this, copulas are proposed for the estimation of the MI, Tsallis and Renyi divergence based information and these are used to evaluate the quality of image fusion.