Multi-focus Image Performance Evaluation Method Based on the Extraction and Combination of Multiple Metrics

Abstract

In multi-focus image fusion, some objective metrics were proposed to evaluate the fused images and compare the corresponding fusion algorithms further. For an image, some metrics can acquire satisfied results, while others can't give credible conclusions. Single objective metric can't assess an image correctly and comprehensively, so multiple metrics may be combined to evaluate the fused images. In this paper, a method named multiple metrics extraction and combination is studied. First, a total set with some popularly cited and representative metrics is formed. Second, for each group of images, the appropriate metrics are extracted from the total set to construct an evaluation set. Therefore two sets of images may have diverse evaluation sets. Third, for each image of the same group, the extracted metrics are combined into three scalars with three measurement methods respectively. Finally, the consistency of measurement results is verified with the scalars. The simulations show that our method outperforms single metric and provides a new way to the research of performance evaluation.