YCbCr subsampling

[MarkT]

Looks like PWP8 (Apr 29.5) is saving jpg files with 4:2:0 (2 2) YCbCr subsampling. Is there a chance, like in PWP7 jpg save options, to offer the choice of susbsampling levels, including 4:4:4?

[jsachs]

It doesn't look like GDI+ supports that option.

[MarkT]

Oh, that's disappointing... so no chance for higher quality jpg saves in future versions of PWP8 then?

Upon further investigation, PWP8 appears to be using 4:2:0 (2 2), taken from ExifTool, which is maybe not the best quality, according to https://www.impulseadventure.com/photo/ ... pling.html.

I'm just curious about future version possibilities.

[jsachs]

The reason JPEG usually does 4:2:2 sampling is that the eye resolves far more detail in luminance (Y) than in chroma (CC), as the article you reference points out. This means you can throw away a little of the chroma information and the difference will be virtually invisible. In my experience, the difference between 4:2:2 and 4:4:4 is insignificant - particularly for photographic images that are rarely if ever sharp at the individual pixel level - compared to setting the quality setting higher or lower. For some computer-generated images that have super-fine detail, you might be able to see the difference if you look at the image zoomed on the screen to 1:1 or higher.

[MarkT]

I hear what you're saying that there may be a slight difference between 4:4:4 and 4:2:2, only PWP8 is producing jpgs with 4:2:0 2x2 subsampling...

Will have to compare some files from PWP7 to PWP8 to really see if I can spot much difference.

[MarkT]

I am pleased to report that after a couple of comparisons, I cannot see any differences in the details between jpg files from PWP7 (4:4:4: 1x1) and PWP8 (4:2:0 2x2).

Started with the same 48-bit tiff file, and saved jpg images out at 100% quality in both PWP7 and PWP8. Zoomed in to 10:1, I cannot see any significant visible differences. Run them through the Difference Transformation and the result is fully black - no difference.

I did find some differences in the files that I find curious. The PWP8 file is slightly larger than PWP7, where I would have expected the 4:4:4 files to be larger... At first I thought the embedded script might account for the difference, but got the same results with or without an embedded script.

Also, I found that the PWP7 files are encoded Little-endian, and PWP8 files are Big-endian. I have no idea what, if any, effect that may have on image quality, but thought it was interesting.

[jsachs]

There is no standard definition for JPEG quality settings, so 90% with one package does not necessarily mean the same thing as 90% with another, so size comparisons at the same quality settings are meaningless. What you need to do is to adjust the quality settings so you get images of about the same size and then compare them visually.

Little-endian vs Big-endian refers to the byte order and has no effect on the image quality. Windows software is generally little-endian which corresponds to Intel byte order. Before Apple switched to Intel processors they favored big-ending byte order which Motorola processors use.