X8 - Chroma subsampling in Jpeg optimiser

[flagpole]

Does anyone know if this long standing bug is fixed yet?

I'm on upgrade strike until they fix it.

(if anyone doesn't' know what this is in versions what X2 through X7 the chroma subsampling options in the jpeg optimizer are basically reversed. No subsampling is maximum, vice versa too. but i don't know about all the options in between)

[MarkZ]

I don't know a lot about subsampling but did some searching and experimenting. I could not find details for the terminology used in PSP.

Using none for the subsampling yields a larger file than with other sumsampling selections.

I experimented with a photo and could see no difference in quality between "none" and the other choices.

I then created a graphic and results are in the images shown below. The lines in the images shown are spaced one pixel apart in one portion and two in another, both in the vertical and the horizontal. The first image was no subsampling and the second was the final selection. The second image is a "Save As" from the first. I had the same result if I created an image and saved it with the same level of subsampling as in the save as. In each illustration the left side shows the image I worked with and the right side is a zoom of what is on the left. The images are much clearer in X8 than what you see here.

No subsampling

Max subsampling

My interpretation is that the subsampling is working correctly with respect to no subsampling vs. some. I have no idea whether the various versions do their jobs properly or not.

I would love to see comments about this from the Corel developers.

[flagpole]

Thanks for that mark. I agree with your interpretation.

bad news is i now have to buy it.

[Jean-Luc]

Thanks for that mark. I agree with your interpretation.

bad news is i now have to buy it.

Just curious : in which case do you use chroma subsampling and different settings of chroma subsampling ?

[MarkZ]

Flagpole:

I did a quick check - Chroma subsampling is correct in PSP X7, version 17.3.0.30. If it was incorrect when first launched it may have been fixed in the SPs. X6 and X5 are backward.

Jean-Luc:

I haven't used this option. Perhaps flagpole can advise. It appears that the file size can be reduced if the loss of detail is not of concern.

[JoeB]

Thanks for that mark. I agree with your interpretation.

bad news is i now have to buy it.

Just curious : in which case do you use chroma subsampling and different settings of chroma subsampling ?

I'm not @flagpole but reading his post I get the impression that he doesn't use chroma subsampling. I say this because his concern is that previous versions of PSP apparently (as per his post) had a bug in that selecting "none" actually invoked max subsampling and selecting maximum actually resulted in no subsampling. As he stated that he had no idea what was happening with the settings in-between I assumed that meant he didn't use them at all.

Personally, and while I am extremely far from being knowledgeable on this subject, I don't see any reason in today's tech age to use chroma subsampling - in most cases - to help reduce image file size when compressing, and what research I've done on the 'net would seem to support that. The main reason for compression used to be for websites but with modern internet speeds and space available to host sites smaller file size isn't nearly important as it used to be. And of course for print purposes we shouldn't be using lossy compression techniques at all.

The only exception would likely be social media sites, forums like this one, etc., which do limit image file sizes that can be uploaded. But in the case of social media it is my humble opinion that it isn't the quality of the uploaded image that most people care about, it is simply sharing the experience, so nobody is going to notice the color shifts caused by using chroma subsampling, and they may be less noticeable in those displays than jpg compression halo artifacts caused by heavy compression settings with no chroma subsampling.

[flagpole]

So chroma subsampling:
If people don't know. You have less detail in the colour than in the bightness (chroma v luma) it makes a lot of sense with real photos because genuinely the eye is able to see less detail in the chroma than in the luma. some, many infact, digital cameras will actually use chroma subsampling on the sensor. even professional grade movie cameras.

It's not a case of these days there is no need to use such a bandwidth saving technique. with real photos they will look better if you use some chroma subsampling and save bandwidth there that you can then use for a lower compression level. it's optimisation, it's why they include it in jpeg, if you ever do optimise a picture for whatever reason, then sometimes that process will involve some chroma subsampling as being the way to get the best picture for a given file size.

(i'm not going to go in to why you would want to optimise at all but i used to run a website getting 5m uniques a day, it makes a difference to your bandwidth bill, server capacity and backup time. it's not just that though even cheap cameras these days can produce 10megabyte images, tricky to cloud-backup. everyone is in a different situation and has different needs, i wouldn't presume to know anyone else's)

If however you wish to annotate a jpeg with graphics you need the subsampling off.

both.png (82.72 KiB) Viewed 9317 times

I did a quick check - Chroma subsampling is correct in PSP X7, version 17.3.0.30. If it was incorrect when first launched it may have been fixed in the SPs. X6 and X5 are backward.

Jean-Luc:

It certainly was wrong at launch. i installed the demo. that is when i went on strike. i didn't notice it in the change log.

As he stated that he had no idea what was happening with the settings in-between I assumed that meant he didn't use them at all.

I meant the first and last entries were the wrong way around. i just meant that i didn't have the time or the inclination to go through all the other entries to see whether the list was upside down or random. it's not as simple as level 0 to 9 it's a matrix of different types.

[JoeB]

I will agree, based on what you said, that there are cases where even with today's available bandwidth and storage a case can be made for compressing some images. And it is useful for video because of the already huge size of video files as compared to image files taken by most people. So I'm basically speaking of compression with respect to image compression in digital photo editors. Whether chroma subsampling is the best choice as opposed to somewhat more aggressive jpg optimization is, however, a bit more complicated IMHO. The 2 links below show examples of observable color changes resulting from using that method (there are other examples available too, of course):

http://www.red.com/learn/red-101/video- ... ubsampling

http://users.wfu.edu/matthews/misc/jpg_ ... maSub.html

So chroma subsampling:
If people don't know. You have less detail in the colour than in the bightness (chroma v luma) it makes a lot of sense with real photos because genuinely the eye is able to see less detail in the chroma than in the luma. some, many infact, digital cameras will actually use chroma subsampling on the sensor. even professional grade movie cameras.

Basically the human eye perceives less detail in color channels than brightness channels, which is why, when viewing an image, a boundary between two colors seems to have less "impact" (for want of a better term) than one where there is a high brightness or contrast difference at a boundary. Chroma subsampling does emphasize brightness or luminance at the cost of color. But, as examples show, this can further reduce the impact of how the eye perceives a boundary between colors because it can cause some color bleeding across that boundary.

And yes, most digital cameras already employ some chroma subsampling (as well as jpeg compression), which means they are already compromising the image stored in the camera to some extent. I guess I'm just not a fan of adding to that compromise any more than necessary. On another note about digital cameras, even without the compression algorithms they use consumer digital cameras simply cannot resolve and capture either the color or brightness of a scene as it is perceived by the human eye because they simply lack the sensors the human eye has as well as the way the human eye changes how it interprets a scene as it scans different color and brightnesses within the scene. That is what the algorithms used in the Athentech Perfectly Clear plugin try to compensate for when that software analyzes a digital image.

It's not a case of these days there is no need to use such a bandwidth saving technique. with real photos they will look better if you use some chroma subsampling and save bandwidth there that you can then use for a lower compression level. it's optimisation, it's why they include it in jpeg, if you ever do optimise a picture for whatever reason, then sometimes that process will involve some chroma subsampling as being the way to get the best picture for a given file size.

As the second link I provided indicates, it does not appear to necessarily be the case that you will get a better compressed image using a combination of chroma subsampling and lower jpg compression than you will sometimes get using no chroma subsampling and a higher jpg compression setting for the same file size. That particular example shows just the opposite to be the case. So perhaps it depends on the image? At the very least, the subject is somewhat more complicated than simply determining that chroma subsampling should necessarily be used.

[flagpole]

The link to red is very interesting. You'd never guess looking at that picture of the eye it's colour had been so aggressively sub sampled.

I'm not saying I always want to use chroma sub sampling. Just that it is a useful tool to have in the arsenal and in particular if the option is there then it should be labeled correctly.

Possibly it's noteworthy that in previous versions of PSP chroma subsampling was part of the default JPEG profile. I don't know if this was a mistake.

[Jean-Luc]

Thank you, JoeB,
the links are very interesting !
And your explanation too...

[JoeB]

Thank you, JoeB,
the links are very interesting !
And your explanation too...

Hope it helps people make decisions as to when and when not to use chroma subsampling. I had another thought but it's just my opinion. That is, I do know that chroma subsampling was originally developed (back in the 1950's I think) as a way to help deal with the issue of huge file sizes with video.. It's intent was not for image files. And it seems to me that it is a reasonable choice for video, NOT just because videos are huge file sizes, but because when you watch video you're mostly watching action in its various forms - e.g., movement of people and other animate objects, watching for facial and body reactions to other influences within the story line, etc. The enhanced contrast (luminance) will help make the scenes "pop" somewhat more, and at the same time people are less likely to be paying any attention to color boundaries in any real detail so the loss of some chroma - and even slight color bleeding - will likely not be noticed. IMHO this is different from how people view a single still image. A still image is a single scene without movement, and people are going to be staring at that single scene for some length of time, unlike they do when viewing a video. And while contrast or luma does add to the impact of that single scene, the length of time it is being viewed allows the eye/brain to also notice the somewhat less color contrast at color boundaries caused by chroma subsampling.

So MHO is that chroma subsampling is best left to being applied to videos and should only be used as a compression method with still images in situations where file size really is a very important and deciding factor.

[MarkZ]

I think you're right JoeB about chroma being used chiefly for video. In searching the internet for information about it I found lots of references to video and very few to images. I did not see the Corel format for chroma subsampling anywhere.