MEF-7 and ICVL Datasets: Evaluating Multi-Exposure Image Fusion Algorithms

This paper utilized two datasets:

1. Due to the lack of publicly available large-scale multi-exposure image fusion datasets, the authors built a dataset called MEF-7. MEF-7 contains 7 sets of multi-exposure image sequences with varying scenes and lighting conditions. Each sequence includes 3-5 short exposures and 3-5 long exposures, totaling 31 multi-exposure images. This dataset aims to provide a publicly accessible test set for evaluating different multi-exposure image fusion algorithms.

2. Additionally, the authors employed a dataset called ICVL, created by Rahman et al. It comprises 34 pairs of multi-exposure image sequences from various scenes and lighting conditions. Each sequence includes 2 short exposures and 2 long exposures. The purpose of this dataset is to assess algorithm generalization ability and performance in real-world settings.