All exports done in Convert & Modify:
Nr KBytes File Variant
0) 747.51 PNG, ARGB, original, ICC "Color LCD", resolution tag "144 PPI"
1) 281.72 WEBP, ARGB, lossless, Exif, ICC "Color LCD" preserved, PPI info preserved
2) 277.47 WEBP, ARGB, lossless, saved entirely w/o metadata, no color conversion done, hence interpreted as sRGB (hence colors a bit off)
3) 289.78 WEBP, ARGB, lossless, perceptual color conversion from "Color LCD" to sRGB (via ☑︎ use batch), Exif, without ICC flag/profile (sRGB default anyhow), PPI info preserved. I expected °3 to be somewhere in the size range between °1 and °2, but it is clearly above both!
So WEBP's lossless compression seems to work somehow less efficient after the pixels raw RGB values got a perceptual color conversion from "Color LCD" to "sRGB". Any ideas why?
My explanation is something like this: Some neighboring pixels (e.g. on continuous surfaces or on boundaries / gradients) which in the old color profile were close or identical enough to be efficiently packed, after the perceptual color conversion were a bit "further apart" which requires a tad more compression bandwidth here and there. Does this seem plausible?
Or is it the alpha channel in particularly which performs significantly worse after color conversion?
Infos appreciated!
[Closed] Lossless WEBP compresses worse after perceptual color conversion from ICC "Color LCD" to sRGB. Any idea why?
- forum_adm
- Site Admin
- Posts: 1819
- Joined: Fri Dec 23, 2016 9:41 am
- Location: Germany
- Country: Germany
- Contact:
Re: Lossless WEBP compresses worse after perceptual color conversion from ICC "Color LCD" to sRGB. Any idea why?
The pixel values are transformed due the color profile conversion.
Normally this cause not much difference.
The saving of the Exif record can occupy the additional 4kb.
Normally this cause not much difference.
The saving of the Exif record can occupy the additional 4kb.
Re: [Closed] Lossless WEBP compresses worse after perceptual color conversion from ICC "Color LCD" to sRGB. Any idea why
Metadata overhead explains the difference between °1 and °2, which is ca 4KB.
But °3 is another extra 12KB which metadata alone cannot explain.
But °3 is another extra 12KB which metadata alone cannot explain.