PSD or jPEG= confused??

[GeoffWessex]

I think you'd have to compare quite large prints before you could really tell a difference (as Matt suggests, really)..... I'd say that, up to around 15x22inches or A3 you wouldn't notice..... A2 may show a difference.

[MattUK]

Originally Posted by chris-p

Thats not strictly true... JPEGs will only discard data if you force them into either generating a new MCU, performing the discrete cosine transformation again or running their colour space conversion again.
In effect, if you open a JPEG and just save it, or save as, using the same packet settings it won't discard data.

Despite the rather odd addition to the conversation due to the unlikelihood of the scenario, if you save a JPG, even at 100% quality, it will lose quality, however imperceptibly. Thus if you open a JPG, then Save, it will lose data.

[KeithT]

Yes, there is a lot of confusion about jpeg files. Opening and closing a jpeg file will not degrade it. If you shoot jpegs only and need to edit it beyond the camera edit, then it is best to save it as a non lossy file before you work on it. You can't add info to it, but you can prevent further loss by saving that file as a tiff, or other non-lossy file format. From that master you can run off as many jpeg files as you like, and providing you don't edit and save the jpegs beyond the initial save-as you won't lose any more information. Shooting jpegs is fine if you first save them as non-lossy for archiving and editing.

You may come to the conclusion, as most of us have, that if you care about the quality of your files you may just as well start shooting raw and making jpegs from your master file when needed.

[OldBoy]

Originally Posted by KeithT

Yes, there is a lot of confusion about jpeg files. Opening and closing a jpeg file will not degrade it. If you shoot jpegs only and need to edit it beyond the camera edit, then it is best to save it as a non lossy file before you work on it. You can't add info to it, but you can prevent further loss by saving that file as a tiff, or other non-lossy file format. From that master you can run off as many jpeg files as you like, and providing you don't edit and save the jpegs beyond the initial save-as you won't lose any more information. Shooting jpegs is fine if you first save them as non-lossy for archiving and editing.

You may come to the conclusion, as most of us have, that if you care about the quality of your files you may just as well start shooting raw and making jpegs from your master file when needed.

Or like me, take Raw plus fine jpeg together.

[KeithT]

Originally Posted by OldBoy

Or like me, take Raw plus fine jpeg together.

I gave up shooting jpeg with raw OldBoy. I found that it just added more time to the workflow, because before I started work on the raw I deleted all the jpegs first, often without even looking at them. I now shoot only raw, and make Tiff masters files at 50mb. When I want the odd jpeg for loading to websites I make them from my master file in seconds.

[chris-p]

Originally Posted by MattUK

Despite the rather odd addition to the conversation due to the unlikelihood of the scenario, if you save a JPG, even at 100% quality, it will lose quality, however imperceptibly. Thus if you open a JPG, then Save, it will lose data.

It's not that odd a contribution as I was trying to correct something you'd said. And you've said it again. Opening and just re-saving a JPEG will not cause any data loss as the algorithm doesn't have to recalculate anything. If you alter the image in any way at all or force the format to recalculate, it will chuck data. But if you don't, it won't. That's how it works!

[MattUK]

Originally Posted by chris-p

It's not that odd a contribution as I was trying to correct something you'd said. And you've said it again. Opening and just re-saving a JPEG will not cause any data loss as the algorithm doesn't have to recalculate anything. If you alter the image in any way at all or force the format to recalculate, it will chuck data. But if you don't, it won't. That's how it works!

I'm intrigued. How would you accomplish this?

Any JPG outputting software I use asks me for a quality setting when I save as JPG. As such, it automatically re-calculates the JPG compression when you save, regardless of whether you changed anything or not.

I'm not being argumentative, I'm just curious

[donoreo]

Originally Posted by MattUK

I'm intrigued. How would you accomplish this?

Any JPG outputting software I use asks me for a quality setting when I save as JPG. As such, it automatically re-calculates the JPG compression when you save, regardless of whether you changed anything or not.

I'm not being argumentative, I'm just curious

That sounds like a bad programming choice in the software. I believe Chris's point is that according to the JPEG standard it should not have to.

[chris-p]

Originally Posted by MattUK

I'm intrigued. How would you accomplish this?

Any JPG outputting software I use asks me for a quality setting when I save as JPG. As such, it automatically re-calculates the JPG compression when you save, regardless of whether you changed anything or not.

I suppose one way to try to describe it is it's almost like the JPEG algorithm is slightly sentient (ooh err) but think complete idiot capable of one thing rather than SkyNet...

If you open a JPEG file and re-save it with the quality at 100 (or compression at it's lowerst, whichever way round you want to think of it) the JPEG algorithm reads the MCUs and sees an intact, functional, unaltered DCT. If it can read and use the original quantisation tables, which it should be able to do if the DCT is functional, it won't re-write them. If you don't re-write the table, you don't scan. If you don't scan, you don't chuck data.

There is a massive caveat to this though...
The quantisation tables used are different for just about every single camera and program you can think of. For the tables to be functional you have to have the RIGHT table.
The practical upshot of this is that if you open a JPEG from your camera in Photoshop and save it, you will get a tiny loss of data (the tables aren't functional as the camera table doesn't match that of Photoshop).
Open the same JPEG file again in Photoshop and just save it and you won't get a loss as you've got the correct format of table for that particular run of the JPEG algorithm...

Originally Posted by MattUK

I'm not being argumentative, I'm just curious

Oh... and there was me looking for another good argument!

[Cathus]

Chris, can you try not to talk in riddles - MCUs, DCTs, quantisation tables???

I for one don't have a clue what you're talking about.