Lossless Image Set Compression Using Animated FLIF

Many image datasets are available on the Internet, contributing to the development of computer vision. While huge datasets are useful for research, they are time-consuming to transfer due to their large data volume. In particular, lossless compression has a worse compression ratio than lossy compression. It is considered that a higher compression ratio can be achieved by encoding multiple images together exploiting the features in the dataset rather than encoding each image individually. In this paper, we propose a new method for efficient lossless compression of image sets by combining a minimum spanning tree (MST) and the Free Lossless Image Format (FLIF). The experimental results show that the compression ratio of the proposed method is better than that of the HEVC-based method. We also show that the compression ratio can be further improved by extending the entropy coder of FLIF, but the effect of the compression ratio improvement depends on the characteristics of the images in the set.