question about non-destructive editing

[Robert Roaldi]

Many packages (RAW developers) offer non-destructive editing (e.g., Bibble for one). But if during the workflow one edits the photo with PS or some other tool that permits uprezzing or perspective corrections, doesn't this violate the non-destructive paradigm?  Those tools create new pixels so you effectively end up with a new photo. Do the RAW developers then create a branch for the new photo that is separate from the original?

[Schewe]

"Non-Destructive Editing" is really a misnomer. It's actually "metadata editing" or more correctly, parametric editing where you don't edit the "pixels" only the settings to the pixels that will be applied to the pixels once they are rendered. If you "render them" to edit in a pixel editor, you can no longer edit the settings but the pixels themselves. One option is to edit as a Smart Object in CS2 or CS3 where the original raw file is embedded and only rendered once the image is flattened.

[Monito]

If you make edits in Canon's DPP program, it keeps the original available and you have the option to revert to the original picture.  However, what I do is limit my edits in DPP to minimal capture sharpening, adjusting white balance, and some brightening in the case of under-exposure.  I batch convert to 16 bit TIFF and do the bulk of post-processing in Photoshop.

[digitaldog]

I fully agree with Jeff about the term. It sounds 'new and sexy' and it has the smell of marketing all over it.

First, there's no such thing as a true non destructive pixel edits. That is, IF you stamp an edit onto pixels, you're changing the numbers and there's data loss. Not a huge, big deal, that's why we edit images. If the term is supposed to mean, you edit a file but don't alter the original, well that's also kind of silly because for one, the original isn't being edited, you're either working on a copy (which we've been able to do since day one in Photoshop) or you're using the original as a data source ala grabbing Raw data to build a rendered pixel based image. So the term is kind of silly.

Even using more robust tools to offset the pixel alterations, tools like Layer Adjustments, Smart Objects, at some point, you need to output the file meaning you either flatten the layers and affect the underlying pixels or you print the file in which case the pixels data is affected as its being sent to the printer.

So, if you alter numbers (pixels), there's data loss. If you leave the original alone, you're not editing that data so how can we call it non destructive editing? Its kind of dumb term I wish, like megapixels or defining scanner resolution using two values (1200x2400) should never have appeared in the first place. But its too late, the term is out there and its just important to know what it really means.

[Monito]

One form of non-destructive editing keeps the original and remembers all the editing steps and applies them in sequence whenever you call up the image.  i think this is how DPP operates since it is basically meant for simple but high quality processing.

[digitaldog]

One form of non-destructive editing keeps the original and remembers all the editing steps and applies them in sequence whenever you call up the image.  i think this is how DPP operates since it is basically meant for simple but high quality processing.
[a href=\"index.php?act=findpost&pid=109482\"][{POST_SNAPBACK}][/a]

But the edits are applied to either the original or a copy of the original right? Once that's done, there's damage to whatever pixels its affecting.

You either alter numbers and apply some degree of 'damage' (a harsh word but somewhat true) or you don't affect the numbers in which case, you haven't edited anything.

[Monito]

But the edits are applied to either the original or a copy of the original right? Once that's done, there's damage to whatever pixels its affecting. You either alter numbers and apply some degree of 'damage' (a harsh word but somewhat true) or you don't affect the numbers in which case, you haven't edited anything.[a href=\"index.php?act=findpost&pid=109493\"][{POST_SNAPBACK}][/a]

You can either theorize or you can check it out with the facts.  Perhaps you don't have a copy, but I directed people to Canon DPP.  If you make edits such as Saturation, sharpness, Picture Style, and even trim tool, it does not destroy any pixels in the original file when you save it.  It simply saves the editing instructions.  The file size is not altered, in fact.  But if you reopen it after it has been edited in DPP, the changes stick.  Further, you can even then Revert to Original Settings.

Thus it is non-destructive editing.

[digitaldog]

You can either theorize or you can check it out with the facts.  Perhaps you don't have a copy, but I directed people to Canon DPP.  If you make edits such as Saturation, sharpness, Picture Style, and even trim tool, it does not destroy any pixels in the original file when you save it.  It simply saves the editing instructions.  The file size is not altered, in fact.  But if you reopen it after it has been edited in DPP, the changes stick.  Further, you can even then Revert to Original Settings.

Thus it is non-destructive editing.
[a href=\"index.php?act=findpost&pid=109612\"][{POST_SNAPBACK}][/a]

It saves the edits, fine. But you HAVE to stamp them at some point no?

Its the same with an Adjustment layer. Its the same with any metadata editing that eventually alters exiting pixels. If you have a pixel with a value of 12/22/128 and at some point you want that to be 22/34/156, you alter those numbers and you undergo some data loss. Its not a big deal but its not non destructive.

Bottom line, you can pay me know or you can pay me later. You either change the numbers or you don't. If you don't, its not an edit (yet) and if you do, you undergo data loss on those pixels.

If there's some why DPP is changing pixel values without data loss, I'm all ears.

[Monito]

DPP is not changing the pixels.  It is saving the edits.  I explained this already.

[digitaldog]

DPP is not changing the pixels.  It is saving the edits.  I explained this already.
[a href=\"index.php?act=findpost&pid=109622\"][{POST_SNAPBACK}][/a]

And I tried to explain that ONCE you stamp the edits, you've introduced data loss. So how is this non-destructive?

You can't print or save out the file until you stamp the edits no?

The only true non destructive editing I'd even consider being real is when you build a rendered pixel based image FROM raw data. The raw is never touched. The pixels are brand new with the edits you specified.

If you have an existing pixel based rendered image, you can't edit without undergoing data loss.

[Monito]

DPP is not changing the pixels.  It is saving the edits.  I explained this already.[a href=\"index.php?act=findpost&pid=109622\"][{POST_SNAPBACK}][/a]

Do the experiment.  You can make edits with DPP on the Raw file that would change every pixel if it were destructive editing (ex. changing saturation) or would delete pixels if it were destructive editing (ex. Trimtool).  But since you can save the changes and then get back all the original pixels, it is non-destructive editing.  Since you don't believe me do the experiment.

[digitaldog]

Do the experiment.  You can make edits with DPP on the Raw file that would change every pixel if it were destructive editing (ex. changing saturation) or would delete pixels if it were destructive editing (ex. Trimtool).  But since you can save the changes and then get back all the original pixels, it is non-destructive editing.  Since you don't believe me do the experiment.
[a href=\"index.php?act=findpost&pid=109626\"][{POST_SNAPBACK}][/a]

Experiment how? The resulting pixels have the same values or new values? If they have new values, they have undergone data loss. If they haven't changed values, you haven't edited anything. I'd like to know your definition of data loss.

As for going back to the original pixels, I can do that in Photoshop using the Undo command, the history command or by deleting adjustment layers before stamping them to the bkgnd pixels. That's not non destructive editing, that's postponing the damage which you have to do if you want to print the file or save it out to any other format (so its non destructive as long as all you do is view it on screen).

What experiment do you propose that proves that original pixels that are numerically altered have undergone no rounding errors (data loss)?

[Pete JF]

Andrew, In a program like Lightzone, which has selection tools in it's RAW software, are the edits made using the selection feature destructive?

I realize that the destruction occurrs after you leave your RAW software and start to process in a pixel editor but I have always wondered why other RAW processing apps haven't tried to incude some basic, featherable, selection tools...was wondering if it was becase they would be "destructive" in some way.


On that note, I wish Lightroom contained some basic selection tools so a person could edit regions before committing to PS. That would be pretty cool...Im sure they left that out for a very good reason though...PS sales?

[Robert Roaldi]

My original question was based on the concept of a non-destructive Bibble-type browser RAW developer for which you use PS (or other) as external editor. (It's easier to write this using actual names of products but the workflow could use other products of course.) This workflow type is commonplace, I believe.

As a follow-up, let me ask this. I am assuming an environment where edits are saved as a sequence of commands that act on an original RAW file. I think that Bibble does this. I can sort of picture how this can be done. The RAW file is left alone and the edit-command sequence is saved in another file. For most kinds of edits, I can conceive how this software works.

But if you call up PS and do an uprez, I don't see how the software can store that editing step as an edit-command sequence operating on the original RAW. The uprez step added NEW pixels and somehow that info has to be saved somewhere. At that point, you have two separate photos it seems to me, the original and the new uprezzed one. I don't believe that the original is overwritten because that would violate the principle of leaving the original always available. How does the browser (Bibble in my example) keep track of the two different-sized images?

Thinking about it, this is more a software design question than anything else.

[digitaldog]

Andrew, In a program like Lightzone, which has selection tools in it's RAW software, are the edits made using the selection feature destructive?
[a href=\"index.php?act=findpost&pid=109635\"][{POST_SNAPBACK}][/a]

No IMHO. This is because all (well almost all) Raw processing isn't altering the original raw data, that's just a source for building pixels. My definition of 'destructive' editing is alter existing rendered pixels.

[aduke]

My original question was based on the concept of a non-destructive Bibble-type browser RAW developer for which you use PS (or other) as external editor. (It's easier to write this using actual names of products but the workflow could use other products of course.) This workflow type is commonplace, I believe.

As a follow-up, let me ask this. I am assuming an environment where edits are saved as a sequence of commands that act on an original RAW file. I think that Bibble does this. I can sort of picture how this can be done. The RAW file is left alone and the edit-command sequence is saved in another file. For most kinds of edits, I can conceive how this software works.

But if you call up PS and do an uprez, I don't see how the software can store that editing step as an edit-command sequence operating on the original RAW. The uprez step added NEW pixels and somehow that info has to be saved somewhere. At that point, you have two separate photos it seems to me, the original and the new uprezzed one. I don't believe that the original is overwritten because that would violate the principle of leaving the original always available. How does the browser (Bibble in my example) keep track of the two different-sized images?

I don't know how Bibble does it, but I would simply remember the target resolution and the interpolation method. Earlier and later "edits" would also be remembered, in sequence. When the image is "stamped", "developed" or "exported", all of the saved edits would be replayed, in sequence, into the resulting file.

Thinking about it, this is more a software design question than anything else.
[a href=\"index.php?act=findpost&pid=109637\"][{POST_SNAPBACK}][/a]