Distilling Style from Image Pairs for Global Forward and Inverse Tone Mapping

Abstract

Figure 1: Distribution of the latent color code representation for different frames. We see that frames with similar content and similar color retouching have been encoded close together in the latent space. All frames are taken from the BBC documentary Blue Planet II Episode "Green Seas". Note that the model trained to encode the style of frames in 2 dimensions (see Sec. 3.4.2) is used in this plot for better visualization.