Jump to content

Lossy test...31 generations of JPEGS compared


Recommended Posts

Do you mean crop and enlarge the image each generation? 

For my own work I settle on the size first and seldom change it. Only burning and dodging usually change, sometimes sharpness and contrast.

Have you done testing on this, if so lets see it.

Link to comment
Share on other sites

In the real world, nobody would simply export the same photo out to a new copy, then do the same to that copy and the subsequent copies through 31 generations, without altering the photo in some way. Scaling might be one thing, but any color adjustments, cropping, added text, watermarking, changes in compression level, etc would alter the image in subtle ways and that will ripple through all subsequent generations. 

JPEG is highly a lossy compression scheme. It's just how it is.

But there's another problem with JPEG: it's only 8 bit, so it's not really suitable for much other than viewing on a computer screen. You wouldn't get as good a print from a color scan of a still photo to JPEG as you would with 16bit TIFF, for example. 10bit is sufficient color depth for scanning most (still or motion) film. Anything less is missing a ton of color information. Anything more than 12 is probably overkill.

Edited by Perry Paolantonio
Link to comment
Share on other sites

8bit is all I've ever used Perry. It works fine for my prints and for digital online. I don't make billboards, but it works for 17 x 22 size and maybe bigger. And in reality I just print 11 x 14 or 13 x 17.

Another problem with going big bit size is space. 16bit is too much for me to work with for storage. I can't afford tape drives or huge RAID and NAS setups Perry. Take the cyanotype I used in the example. The original 800 dpi TIFF scan cropped was 97mb. The uncropped version full bleed to edges was 151mb in a 24 bit color scan. For the same scan in 48 bit color it is 303mb. That size of file is too big for me. I got hundreds of thousands of images, I could never begin to store those size files on my budget. And some of my scans are a lot higher dpi than 800. 

There is an interesting difference if you go the Internet Archive link and look at both low res images without clicking on them. There are 2 small dots at the bottom of the image. Click on the dots and it toggles between the gen 1 and gen 31 images. The 31st gen jpeg has lines in it. But when you click on the photo and it brings up the hi res version there are no lines. Maybe if it was printed in 40 x 60 inches the 31 gen file may have lines in it...who knows? 

What is your explanation of this phenomena?

 

Edited by Daniel D. Teoli Jr.
Link to comment
Share on other sites

I don't know what to say here. Hard drives are cheap. You're definitely missing out on a lot of data at 8bit. This is objectively, empirically proveable. An 8-bit image may look ok to the human eye, but the minute you start manipulating the images, you will see it fall apart. 

Quote

What is your explanation of this phenomena?

It's a scaling artifact. Because the 1st and 31st are probably subtly different from all those generations of compression. 

Link to comment
Share on other sites

16 hours ago, Perry Paolantonio said:

I don't know what to say here. Hard drives are cheap. You're definitely missing out on a lot of data at 8bit. This is objectively, empirically proveable. An 8-bit image may look ok to the human eye, but the minute you start manipulating the images, you will see it fall apart. 

It's a scaling artifact. Because the 1st and 31st are probably subtly different from all those generations of compression. 

Well, I work in many areas Perry and am limited in what I can spend in any one area. I have to try to balance my budget. I have a number of hard drives already. But they add up in $$. You have working drives, back-up drives, off-site drives, etc. I have to budget to buy postage stamps when I send out promotional mailers, so can't just spend on whatever I want to.  

Link to comment
Share on other sites

To add some manipulations I extended the test to 32 generations, reducing and then enlarging the 32nd gen JPEG. If you manipulate an image enough you will be successful in degradation. But this test was done simply to show how importing and exporting a JPEG affected the image over many generations.

Link to comment
Share on other sites

33 minutes ago, Daniel D. Teoli Jr. said:

To add some manipulations I extended the test to 32 generations, reducing and then enlarging the 32nd gen JPEG. If you manipulate an image enough you will be successful in degradation. But this test was done simply to show how importing and exporting a JPEG affected the image over many generations.

Understood, but my point was that this is not a really useful test in the first place. I mean, why would one open a file and save as to a new JPEG of exactly the same size, crop, color, etc? Why not just make a copy of the file? A more realistic scenario would involve adjustments to the master image, saved to a new file, then further adjustments to that. 

Translating that to the motion picture world, you'd make a master, then you'd make adjustments to derivative files made off that master. For example - You have a 2k copy of a flat scan of a film (Generation 1). This gets color corrected and rendered to a new copy (Gen 2), and that gets rendered out and brought into a restoration system (Gen 3). That's then put into a final deliverable master and rendered out as the Restored Master (Gen 4). Now you need an HD file so you'd render a downconvert to a new file (gen 5). Then someone asks for a texted version with lower thirds, logos and other stuff, and you're at generation #6. If you did this in a format like Uncompressed, or a high end ProRes (4444, XQ  or HQ) or Avid DnXHR or similar, your footage will hold up through many generations without appreciable loss. If you did this with a lossy format like JPEG, it definitely will not.

This is not an unusual workflow either. We literally do this 5-6 times per month on feature length films. 

Link to comment
Share on other sites

  • Premium Member

Bear in mind that ProRes, of any stripe, is essentially the same technology as JPEG. It's a little refined, so the quality to size ratio is slightly better, but a (suitably high bit rate) JPEG and a ProRes frame have pretty much the same characteristics.

P

Link to comment
Share on other sites

14 minutes ago, Phil Rhodes said:

Bear in mind that ProRes, of any stripe, is essentially the same technology as JPEG. It's a little refined, so the quality to size ratio is slightly better, but a (suitably high bit rate) JPEG and a ProRes frame have pretty much the same characteristics.

Compression-wise, yes. ProRes uses a similar intra-frame compression scheme with a pretty low compression ratio. However, ProRes is 10 (or 12) bit unlike JPEG's 8bit, so it stands up to a lot more before you start to see problems. 

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...