A new Image Similarity Metric for a Perceptual and Transparent Geometric and Chromatic Assessment

The authors propose EDOKS, a new perceptual similarity metric combining two terms: Earth Mover's Distance for texture dissimilarity and Oklab space distance for chromatic dissimilarity. This metric outperforms existing state-of-the-art methods, especially with shape distortions, and offers visual explanations to justify its scores, making similarity assessment transparent.

Open the full episode

More from AI Radio