PearsonZero

3 ms·

1. 1.▲
   Per-Image BT.601 Decorrelation Gap Measured Against KLT Across the Kodak Suite(github.com)

   1 points by PearsonZero 5mo ago|0 comments
2. 2.▲
   by PearsonZero 5mo ago

   Code is now in the repo — ‘’’pip install numpy Pillow’’’
3. 3.▲
   by PearsonZero 5mo ago

   Code is now in the repo — ‘’’pip install numpy Pillow’’’ where you can duplicate the values. Kodim might seem outdated, but it’s still the primary benchmark cited in learned image compression research (CLIC, neural codec papers) and is refe
4. 4.▲
   Per-image PCA characterization of the Kodak image suite (PDF and JSON)(github.com)

   7 points by PearsonZero 6mo ago|4 comments
5. 5.▲
   First per-image PCA decomposition of Kodak suite reveals deliberate curation(github.com)

   9 points by PearsonZero 6mo ago|8 comments
6. 6.▲
   Channel decorrelation: 52.8% reduction across Kodak suite, no ML or codec mods [pdf](github.com)

   1 points by PearsonZero 6mo ago|0 comments
7. 7.▲
   Inter-Channel Decorrelation Below R=0.01 with Spatial Autocorrelation Above 0.99 [pdf](github.com)

   1 points by PearsonZero 6mo ago|0 comments