First per-image PCA decomposition of Kodak suite reveals deliberate curation
Summary
The Kodak PCD0992 Statistical Profile Series presents the first per-image PCA and inter-channel redundancy analysis for all 24 images, detailing covariance, eigendecomposition, and derived metrics. It provides a reproducible data-characterization workflow with reference data sheets and JSON metrics, highlighting deliberate curation across the image suite and implications for understanding image-data diversity in AI datasets.