PDA

View Full Version : A question for the digital experts here



chris jordan
13-Nov-2006, 13:31
Hi guys, here's a question about the quality of JPEG's, for those who really understand this stuff. I have several very large image files that I am having printed at Duggal. To save the image as a TIFF requires using a DVD because they're 800MB, larger than what a CD will hold. But if I save the images as 12-quality JPEG's, they are only 140 MB, which can easily be stored on a CD.

So out of curiosity, I tried saving the file both ways, and then opened both files and compared them at 1600%. I could see no visible difference, even in the areas of highest detail and tonal smoothness. The JPEG had every bit of detail and color as the TIFF file did. So the question is, is there any real difference between saving files as uncompressed TIFF's (which takes tons of space), and using the highest-quality JPEG option? If 12-quality JPEG's are just as good, I could save the most humgous amount of hard drive space here.

~cj

PViapiano
13-Nov-2006, 13:41
Hi Chris...

While I've never compared as you did, I always send my digital files as 12-quality JPEGs and I've never seen any discrepancies/differences at all in the finished work.

However, I always archive my master files as PSD docs, with all layers intact.

Scott Rosenberg
13-Nov-2006, 13:42
chris, i'm not an expert by any means, but when comparing the images on a screen, i would imagine that the limitations of the display would outweigh the differences between the format the file is saved under. have you tried making prints to compare? someone here might have already done this, but i suspect that would be a better way of comparing quality sacrificed by jpeg compression.

PViapiano
13-Nov-2006, 13:46
i would imagine that the limitations of the display would outweigh the differences between the format the file is saved under

Hmmm...viewing at 1600%, if there are no differences in color or contrast, I'd say you'll never see it in a print of any practical enlargement.

Ted Harris
13-Nov-2006, 13:47
Scott is right about the limitation s of some screens. Beyond that .... TIFF files are lossless meaning that you can open them again and again and never loose any of the information. JPEG's OTOH are losey meaning that every time you spen and then resave a JPEG you lose a tiny bit of information, after some number of iterations of opening and closing (resaving) there will be some image degradation. Much of the degradation n a JPEG is due to the fact that the info is always stored in a compressed format ... even if you save at the highest value. TIFF's OTOH are saved uncompressed unless you specifically request they be compressed and then you have compression choices that will dramatically reduce info loss, choices that are not available for JPEG's.

BTW, if you are concerned about how you store the data for printing why not ask if they have an option for you to use an FTP client to upload your files?

Greg Miller
13-Nov-2006, 13:51
The biggest difference between yopur master image and a 1st generation 12-quality jpg would be if your master file is 16 bit. jpg's are by default 8 bit files. If your master file is also 8 bit then ther will not be any tone shifts in the jpg.

The amount of lossy compression in a 12-quality jpg is quite small so I doubt you would be able to see any compression artificats in your print.

Greg Miller
13-Nov-2006, 13:56
The biggest difference between yopur master image and a 1st generation 12-quality jpg would be if your master file is 16 bit. jpg's are by default 8 bit files. If your master file is also 8 bit then ther will not be any tone shifts in the jpg.

The amount of lossy compression in a 12-quality jpg is quite small so I doubt you would be able to see any compression artificats in your print.



I need to amend my comment on tonal differences. Jpgs also need to be in srgb color space. So if you are converting from a color space with a wider gamut (such as Adobe RGB 1998 or ProPhoto RGB) then you may see a change in tones.

External hard drives are a good alternative here. You can pruchase a very small external hard drive at a pretty cheap $ anout. You could copy your tiff file to the external hgard drive and then ship that to Duggal.

Greg Miller
13-Nov-2006, 13:59
You should also try using LZW compression of your TIFF files. LZW is a lossless compression. Since you are starting at 800mb, this could very well get you down to a file size small enough to fit on a CD. And with no image degradation worries.

Michael Mutmansky
13-Nov-2006, 14:04
...is there any real difference between saving files as uncompressed TIFF's (which takes tons of space), and using the highest-quality JPEG option? ...

~cj

Chris,

Have you tried using compressed TIFF format? The LHZ or ZIP compression in TIFF is a lossless compression method. It doesn't look at the image, it looks are the strings of bits and does compression where it finds data that can be compressed. That may reduce your size to CD levels.

As Ted says, JPEG is a 'lossy' compression method, but I wouldn't hesitate to use the highest quality setting for a file that I need to send out. Just avoid ever having to open and re-save that file because the compression artifacts will increase when this is done. Generally, I'll make a JPG for that purpose, and then dump it so it doesn't get in the editing pipeline again. I'll go back to the TIFF or PS file (with layers) for more editing.


---Michael

Michael Mutmansky
13-Nov-2006, 14:08
Greg,

JPG is sRGB only if you don't specifically assign a profile. It can have any normal profile embedded within it so there isn't a loss of color information using JPG.


---Michael

tim atherton
13-Nov-2006, 14:10
Jpgs also need to be in srgb color space. .

Okay - that's a new one on me...?

Chirs, what's the particular problem with using DVD's

Scott Rosenberg
13-Nov-2006, 14:18
Hmmm...viewing at 1600%, if there are no differences in color or contrast, I'd say you'll never see it in a print of any practical enlargement.

Limitations in the screen are present and consistent whether you are viewing at 1% or 1,000,000%. let's say you are trying to measure two lengths of pipe to see which one is closer to your target of 12.125 inches. if your ruler will only give you resolution to 1/10" of an inch, then your gauge will not be capable of returning a reliable result. you need a more precise instrument.

i was suggesting that in chris's case the resolution of the display might be inadequate to show differences in the file. however, a print, which has much greater fidelity than most monitors, may prove to be a more capable 'ruler'.

Daniel Geiger
13-Nov-2006, 14:34
I think the LZW compression should get you there easily, and with the least amount of worry.

If you are really concerned about quality, just burn that DVD. However, some DVDs I burnt on my G5 at home and sent to my stock agency, they had trouble opening. So cross-system/platform compatibility with DVDs is still a bit iffy [my experience so far]. Otherwise, think about a memory stick. The 1GB are not that expensive and re-usable.

For what it is worth, my stock agency now accepts (an soon will ONLY accept!) highest quality jpegs (no more tifs). This may be a question of "good enough", a debate that needs not be revisited. I still archive my files as tifs (on DVDs and a RAID5 array).

Walter Calahan
13-Nov-2006, 15:57
Chris

Save your master copy as a TIFF. Send Duggal Jpegs at #12 quality. As long as they are NOT opening and closing them a gazillion times before printing, you should see no difference. Just don't use the sRGB profile. Adobe 1998 RGB is standard for printing.

Never trust you monitor!

Personally I'd send a DVD, they are not that expensive compared to a regular CD, just your computer system can't burn DVDs.

Ken Lee
13-Nov-2006, 16:18
The differences you see, depend on the nature of the image.

JPG compression replaces adjacent tones of the similar shade, with larger square "tiles". Greater compression means that the algorithm is more tolerant when it comes to chosing an adjacent tone that is almost the same color - so there are fewer, larger tiles. This, as Ted has pointed out, it is a lossful method. Once the similar shades have been tossed out, they cannot be retrieved.

When your original contains large uniform areas (for example a grey sky), such areas can be replaced by a small number of tiles, with little visible impact. On the other hand, if your image contains a lot of small details (such as a sandy beach in strongly angular light) then the replacement will be more obvious, and it will be very easy to distinguish between lossless and lossful compression.

Ken Lee
13-Nov-2006, 16:22
Photoshop has its own lossless compression algorithm. It is not universal, but due to the popularity of PS, it might as well be.

Although non-Mac web browsers can't handle them, you should attach an ICC profile to your original, so that you preserve the color space in which the image is to be viewed. On a calibrated system, this will guarantee that others will see what you intended.

Does anyone know if the new Internet Explorer 7 supports embedded ICC color profiles ?

Jack Flesher
13-Nov-2006, 16:33
Chris, there are several not readily apparent issues to consider...

1) jpegs by default are 8-bit and should be in the sRGB colorspace. sRGB is a very small colorspace and using a larger colorspace can cause banding in even-toned areas of the image

2) you did not see a difference between your jpeg and tiff images because your monitor is also essentially an sRGB colorspace device and cannot effectively display colors outside sRGB.

3) If you instead get the print profile from your lab, then compare the images in CS using CS's "soft-proof" engine, you will likely see a difference, but

4) Unfortunately it takes a trained eye to understand how the relatively minor soft proof differences that show on your monitor translate to significantly larger differences in the print

5) Most all current printers are capable of printing colors far outside the sRGB colorspace, so you will indeed lose some of the available color if you use jpeg sRGB

6) A 16-bit tiff allows a lot of room for your lab to "adjust" and convert your original image to their working colorspace without deteriorating your original image; an 8-bit sRGB jpeg allows essentially none.

7) My .02: In conclusion, if these are good quality, fine art images, you would be well advised to swallow the slight additional cost of burning your tiffs to DVDs and sending your lab the ultimate print file you can send ;)


Cheers,

chris jordan
13-Nov-2006, 17:14
With 12-quality JPEG's I can see no tiling, and it is easy to embed a colorspace other than sRGB so the images stay in the proper colorspace. But the files come out so much smaller that it's hard to believe they aren't compromised in some way.

Scott Rosenberg
13-Nov-2006, 17:30
With 12-quality JPEG's I can see no tiling, and it is easy to embed a colorspace other than sRGB so the images stay in the proper colorspace. But the files come out so much smaller that it's hard to believe they aren't compromised in some way.

chris, you can only see what your monitor is capable of displaying. making a few prints might bring into relief subtleties lost on your display.

Let’s say you and I both hit a golf ball, yours went 325 yards, mine went 305 yards. If our judge could only see a distance of 250 yards, would he be able to tell who drove the ball farther? The differences in the files might be outside of the capability of your display. Just because you don’t see them, doesn’t mean they aren’t there.

Jack Flesher
13-Nov-2006, 19:03
With 12-quality JPEG's I can see no tiling, and it is easy to embed a colorspace other than sRGB so the images stay in the proper colorspace. But the files come out so much smaller that it's hard to believe they aren't compromised in some way.

That's because they are compromised -- a quality 12 jpeg is still compressed and jpeg compression is lossy...

QT Luong
13-Nov-2006, 19:40
Personally, I do not see any visible difference with my eye. However, when I run Norman Koren's imatest (a program that can measure MTF in photographs of high-contrast slanted edges) on the same image saved as TIFF and JPEG, the TIFF yields a higher number.

chris jordan
13-Nov-2006, 20:00
If the colorspace of the image is preserved in JPEG compression (which it is), then how would monitor gamut limits come into play? I don't get that part.

Jack Flesher
13-Nov-2006, 20:49
If the colorspace of the image is preserved in JPEG compression (which it is), then how would monitor gamut limits come into play? I don't get that part.

The monitor gamut limit is SMALLER than the colorspace you are using. Hence, it can only show you what it can render, and a normal monitor can only render up to sRGB. ANY color that falls outside the sRGB colorspace -- also the monitor colorspace -- simply isn't going to be seen on that monitor. Moreover, it will usually get rendered as the last available nearest color instead of the real color. For expample, a very bright blue will fall well outside sRGB and will be seen on your monitor as a normal, middle-toned blue.

CS is a colorspace aware program and it knows (numerically) that Adobe RGB is larger than sRGB because of the absolute color number values. Hence, it can use those numbers and scale the visual outputs RELATIVE to each other so it can then render them, again relatively, while remaining inside sRGB (the only colorspace your monitor is capable of) so you can compare them visually inside CS. But this rendering is *not* an absolute rendering, only a relative visual 'estimate' of the differences.

Make sense?

Greg Miller
14-Nov-2006, 04:50
Greg,

JPG is sRGB only if you don't specifically assign a profile. It can have any normal profile embedded within it so there isn't a loss of color information using JPG.


---Michael

Brain cramp. I was confusing my own personal practice/habit with reality.

Michael Mutmansky
14-Nov-2006, 07:34
Greg,

Most people properly leave a JPG untagged when using the file for the internet, so it is my practice to convert to sRGB and leave the JPG untagged for that purpose. I expect you do something similar.


---Michael

Leonard Evens
14-Nov-2006, 07:46
If you save a file once for a specific purpose as a jpeg, you aren't going to see much loss of quality. But you shouldn't use jpeg for routine storage, particularly for images you are going to be editing.

I use png format, which is also a lossless format, and it allows higher degrees of compression thatn I've found available for tiff. You might look into it.

Given the increases in computer speed and memory, increasing resolution of digital sources, I think it is inevitable that DVD and other media for storing gigabytes inexpensively are going to become the standards for storage.

chris jordan
14-Nov-2006, 07:46
Jack, I understand that monitors can't show the whole gamut of a photograph. But here's the part I don't get. My monitor can't show the full gamut of the original Photoshop file either, which is in Ektaspace. If I convert that file to a 12-quality JPEG, the monitor still can't show the full gamut. But that doesn't prove that the image has degraded in quality, and considering that the image is still in Ektaspace as a JPEG, there is no reason to think any of the colors have been clipped.

What would be really cool is if someone who understands how the JPEG algorithm works would show up here. There must be such people out there (but I'm betting they're not into LF photography...).

~cj

Henry Ambrose
14-Nov-2006, 08:02
Here Chris,

Go read this. http://tinyurl.com/nrvm9 It does not answer your jpg question but it'll help you understand about color spaces and gamut.

jpeg is not a color space. Its not sRGB or Adobe 98 or anything else color wise. jpeg is a compression scheme.

Jack is sorta right about monitors and sRGB but not precisely right. To know for sure you'll have to verify your monitor's ability to render color. sRGB is the standard for low end office monitors. We don't know what you have, but if you have a better monitor you may be seeing more color than sRGB. There are monitors that do indeed fully render Adobe 98.

By the way, Adobe 98 may not be the end all color space. Depending on what printer you're using another space might be better. Notice I say "might" - You also need to know the gamunt of the printing device being used. Do you know what kind of printer you're sending to? It could be that its gamut is within your monitor gamut and your selected color space, in which case you are good to go. Big inkjets don't have huge gamuts. Pigment inks generally have even smaller gamuts.

Really, if your goal is to make the very best prints you'd start with the capability of the printing device in mind and tailor your system to fit that final output. But this is rarely the case as you probably use more than one output device and each has its own requirements. So picking a color space and set up is usually done by settling on something that works well with various output methods. Adobe 98 is pretty popular becasue it does work well for most uses. Then there's the part about you printing something in 10 years from a limited color space on the latest super printer we have not imagined. If you threw away some color info then you might be scanning your film again. But don't fret about this too much.

If you find that your printer has a gamut that is outside of your current set up (I doubt this) you might want to change how you work or maybe use soft proofing in Photoshop to get an idea of what's gonna happen to your file when printed.

One more thing to consider is that Adobe 98 is really about getting the process color gamut tied to computer displays and color spaces. Its made to do a very good job for printing via printing presses and 4 color inks. It just so happens that it does a good job as a general purpose color space for photography. You could use something else if it was of benefit. (Probably not as Adobe 98 pretty well covers what we do) You'd know it would be a benefit if you found a printer that made colors that you couldn't output from your system in the color set up you now use.

Think of color management as a chain like that of your tripod, camera, lens, development, enlarging or scanner lens and so on. All the pieces add together to make a chain and can be limited by the weakest link. Making each link optimized will give you the best final result.

Finally, I suggest you use either native PS files or LZW tifs - either will save you some space, probably enough to get your big files onto CDs. Why lose anything?

Jack Flesher
14-Nov-2006, 08:04
Jack, I understand that monitors can't show the whole gamut of a photograph. But here's the part I don't get. My monitor can't show the full gamut of the original Photoshop file either, which is in Ektaspace. If I convert that file to a 12-quality JPEG, the monitor still can't show the full gamut. But that doesn't prove that the image has degraded in quality, and considering that the image is still in Ektaspace as a JPEG, there is no reason to think any of the colors have been clipped.

What would be really cool is if someone who understands how the JPEG algorithm works would show up here. There must be such people out there (but I'm betting they're not into LF photography...).

~cj


Chris, I'll try one last time: You have to view the two images side by side using CS's "Soft Proof" tool to see the differences.

If you use that tool properly and if you know what to look for as it isn't always obvious, you will see differences. Understanding how those relatively minor-looking soft proof differences will translate out on the print is a different story and requires some experience to properly evaluate it...

chris jordan
14-Nov-2006, 08:43
But if the gamut limits don't show up in the first image due to the limitations of the monitor, how are they going to show up as something different in the JPEG, which is in the same colorspace? I think it is you who is not quite getting this. In any event, what I REALLY want to know is how the JPEG algorithm works. I am a sophisticated digital printer working in a color-managed environment, and I can see no difference either in print or on my screen at 1600%, between an 8-bit TIFF and that same file saved as a 12-quality JPEG. But the filesize difference is huge, which is strange and interesting.

adrian tyler
14-Nov-2006, 09:01
hi chris, my files often go over 800 mb and i just write as a tiff onto a dvd.

sending you your prints off to you today, off to paris with dely maņana... all the best

ade

paulr
14-Nov-2006, 09:03
In any event, what I REALLY want to know is how the JPEG algorithm works. I am a sophisticated digital printer working in a color-managed environment, and I can see no difference either in print or on my screen at 1600%, between an 8-bit TIFF and that same file saved as a 12-quality JPEG. But the filesize difference is huge, which is strange and interesting.

Here's a pretty in-depth explanation. take a deep breath ...
http://en.wikipedia.org/wiki/JPEG

FWIW, i can't see the difference between 12-quality jpegs and uncompressed files either (on screen or in print). though i haven't done any serious comparisons. To be on the safe side, I don't use any jpeg compression on the digital files I print from. I print much smaller than you do, so I haven't felt much pressure to experimentt with it yet.

adrian tyler
14-Nov-2006, 09:14
yeah, i wouldn't risk it, after all the work one put's into getting it right, just burn a dvd...

Jack Flesher
14-Nov-2006, 09:26
I think it is you who is not quite getting this.

Then I suggest you use jpeg and not listen to me or anybody else...

As for how jpeg compresses, it is the topic of a lengthy white paper. The simple answer is the human visual system perceives brightness differences much more readily than absolute color differences and jpeg compression takes advantage of this fact. It selects a range of 'similar' colors of close to equal luminance and adjusts them all to one color of the same luminance -- and by so doing allows the total image data from an original file to be significantly truncated or compressed. How 'similar' the colors are that it initially selects and how variable the luminance is determined by the compression factor chosen. How any given human imaging system reacts to them is variable...

Best of luck,

chris jordan
14-Nov-2006, 09:58
Well I just read the in-depth article about JPEG's in Wikipedia, along with some of the links (which are even better), and learned a few things. First, as a few people here have suggested, the 12-quality JPEG in Photoshop is lossy, despite not appearing to be so. But the articles made a distinction between "true" lossless compression and "virtually indistinguishable" compression. 12-quality JPEGs are virtually indistinguishable, but not truly lossless. There is a truly lossless JPEG generator out there, but it is hard to find and the file sizes end up being almost the same as a TIFF.

My 30-saves test confirms the above: there is a visible change between the first and 30th versions when viewed at 800%, but it is not a change that appears to my eye to matter. Some individual pixels do shift slightly, but the overall colors stay visibly the same and all image details are preserved.

For printing purposes (and I just tested this too), there is no perceptible difference in prints made from Photoshop files versus 12-quality JPEGs saved from the same original files.

For emergency backup purposes, I think I'm going to do a "save-as" of all of my files to 12-quality JPEGs. That way I can save my whole image library on a single hard drive, which for me is a huge administrative gain. But I'll still print from the original Photoshop files, just because...

Thanks for all your thoughts on this stuff guys, especially Jack.

cheers,

~cj

QT Luong
14-Nov-2006, 10:47
For emergency backup purposes, I think I'm going to do a "save-as" of all of my files to 12-quality JPEGs. That way I can save my whole image library on a single hard drive, which for me is a huge administrative gain.


That's what I do to have access to all my files while travelling.

Jack Flesher
14-Nov-2006, 11:06
At the last sale, I bought Seagate 500G SATA drives for $189 from my local Fry's computer store. That's 38 cents per Gig now and they're only getting cheaper... From my perspective, at that cost, I can afford dual redundant back-up (Raid 1-1, with the second mirrored set stored offsite) onto said drives and never have to worry about catastophic file loss.

And I realize others views will vary.

Cheers,

Greg Miller
14-Nov-2006, 11:13
Greg,

Most people properly leave a JPG untagged when using the file for the internet, so it is my practice to convert to sRGB and leave the JPG untagged for that purpose. I expect you do something similar.


---Michael

Yep - I do the same.