MFFW: A new dataset for multi-focus image fusion

• DOI: 10.48550/arxiv.2002.04780
• Publication Date: 2020-01-01
• Authors: Shuang Xu; Xiaoli Wei; Chunxia Zhang; Junmin Liu; Jiangshe Zhang
• Keywords: FOS: Computer and information sciences; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology; Computer Science - Multimedia; Multimedia (cs.MM)

Multi-focus image fusion (MFF) is a fundamental task in the field of computational photography. Current methods have achieved significant performance improvement. It found that current are evaluated on simulated sets or Lytro dataset. Recently, growing number researchers pay attention to defocus spread effect, phenomenon real-world multi-focus images. Nonetheless, effect not obvious datasets, where popular perform very similar. To compare their images with this paper constructs new dataset called MFF wild (MFFW). contains 19 pairs collected Internet. We register all source images, and provide focus maps reference for part pairs. Compared dataset, MFFW significantly suffer from effect. In addition, scenes more complex. The experiments demonstrate most state-of-the-art cannot robustly generate satisfactory can be baseline test whether an MMF algorithm able deal