Abstract

Multisensor signal-level image fusion has attracted considerable  research attention recently. Whereas it is relatively straightforward to obtain a  fused image, e.g. a simple but crude method is to average the input signals,  assessing the performance of fusion algorithms is much harder in practice. This  is particularly true in widespread “fusion for display ” applications where  multisensor images are fused and the resulting image is presented to a human  operator. As recent studies have shown, the most direct and reliable image  fusion evaluation method, subjective tests with a representative sample of  potential users are expensive in terms of both time/effort and equipment  required. This paper presents an investigation into the application of the Visible  signal Differences Prediction modelling, to the objective evaluation of the  performance of fusion algorithms. Thus given a pair of input images and a  resulting fused image, the Visual Difference Prediction process evaluates the  probability that a signal difference between each of the inputs and the fused  image can be detected by the human visual system. The resulting probability  maps are used to form objective fusion performance metrics and are also  integrated with more complex fusion performance measures. Experimental  results indicate that the inclusion of visible differences information in fusion  assessment yields metrics whose accuracy, with reference to subjective results,  is superior to that obtained from the state of the art objective fusion  performance measures.