An observation on Raw vs Jpeg

I have read posts that say they can't see a difference between Raw and Jpeg. For the G9 Raw, I have noticed that if the Raw has the same settings that the Jpeg was created from and the pictures haven't been post processed then really nothing is different since the rendering engine to your own eyes is only doing 256 states per color. But if you mess with them ...

Each of the 256 states are represented with an integer. When you change contrast, brightness, or any non-linear process (i.e. you don't do the exact same thing to each brightness step), then you immediately get into fractions i.e. levels in between the 256 states. You can't represent fractions with integers so they are truncated to the nearest integer. This is why you start to see posterization in the histograms long before you can see it with your eyes. You can see some of the 256 states starting to collect more integers and some that aren't just because every operation is doing a little chopping. Note that, the actual chopping/truncation doesn't really happen until you save them out in a format like Jpeg that doesn't have fractional representation. So the histograms are showing what is going to happen when you tie back to 256 states per color.

This is why 16 bit format is so much better for editing ... you have 256 additional states in between every normal Jpeg state. Unfortunately when you started with only 256 states Jpeg you really don't have any real data at these other points so you don't receive the full benefit but ...

When you have Raw (12 bit), you have 16 additional states in between each Jpeg state which helps a significant bunch as many have noticed. As long as you preserve all the data going into Photoshop's larger 16 bit math, you have photos that can resist degradation while you do your curves magic on them. When these photos are converted down to 256 value Jpegs, they look better than the same manipulations done to an 8-bit jpeg from the start.

This fact is independent from the other gotchas of not having sharpening, contrasts, and exposure decisions made by the camera before you ever get the image, a la Jpeg capture. All of these in-camera decisions are not easily reversible once the data has been passed to you at only 256 values per color.
John

John Prichard said:

I have read posts that say they can't see a difference between Raw
and Jpeg. For the G9 Raw, I have noticed that if the Raw has the same
settings that the Jpeg was created from and the pictures haven't been
post processed then really nothing is different since the rendering
engine to your own eyes is only doing 256 states per color. But if
you mess with them ...

Click to expand...

Sort of a no-brainer, John. It's all in knowing how to "mess with them"

--
Phil .. Pany FZ18, SD700IS, EOS 300D & 20D; Canon S3 IS; Fuji F20 & F31fd.
http://www.pbase.com/phil_wheeler
http://www.flickr.com/photos/phil_ox/

But when you open your JPEG the first thing you can do is change the color depth to 16 bit.

The real benefit of RAW is the operations that occur at the sensor data level...such as exposure compensation. Image manipulation (no matter if it's JPEG or TIFF at 16 bit or whatever) can't provide the same functionality.

For example, recently in another thread it was clearly demonstrated that a RAW exposure compensation of +4 makes the same kind of adjustment to the RAW converted image as having originally shot the image with an ISO 4 stops higher. That is to say, an ISO 100 capture with an EC of +4 looks nearly the same as an ISO capture of ISO 1600 at the same shutter and aperture. Such adjustments are impossible with image formats because the adjustment is being made to the sensor data before it is turned into an image.

Do you happen to remember the name of that thread? General rule of thumb is that RAW files are only good for about 1.6 to 2 stops at best, so I would be interested in seeing the argument for 4 stops.

fishcatcher said:

Do you happen to remember the name of that thread? General rule of
thumb is that RAW files are only good for about 1.6 to 2 stops at
best, so I would be interested in seeing the argument for 4 stops.

Click to expand...

Yes, that's what others said which, of course, created a lot of back and forth posts. The link to the test shots are around the middle of page 4, with conclusions on page 5. I get into it on page 2 with comments and links to some test shots.

Here's the thread
http://forums.dpreview.com/forums/readflat.asp?forum=1002&thread=26236419

Here's the link to greg's test shots
http://gregorywilson.smugmug.com/gallery/4092096/1/238516772#238516772

As you already know, changing to a color depth of 16 bits after you have already read in 8 bits doesn't do you much good since you have no data to put into those 256 levels between each 8 bit level. So they are all the same shade. The only thing that buys you in an after-the-fact case is not acquiring too much round-off error in further processing.

All the people in this thread are already knowledgeable about what you give up by accepting the 8-bit version from your camera.

I was hoping to accomplish something else ... convince people who see no difference to take another think about it.

Guess that won't happen. Oh well.

John

John Prichard said:

As you already know, changing to a color depth of 16 bits after you
have already read in 8 bits doesn't do you much good since you have
no data to put into those 256 levels between each 8 bit level. So
they are all the same shade. The only thing that buys you in an
after-the-fact case is not acquiring too much round-off error in
further processing.

Click to expand...

After converting to 16 bits, operations performed have the depth necessary to minimize post-processing effects such as posterization, which would have been worse at 8 bits.

John Prichard said:

All the people in this thread are already knowledgeable about what
you give up by accepting the 8-bit version from your camera.

I was hoping to accomplish something else ... convince people who see
no difference to take another think about it.

Guess that won't happen. Oh well.

Click to expand...

Maybe you need a different approach. See the third post, second page in the thread listed above.

http://forums.dpreview.com/forums/readflat.asp?forum=1002&thread=26236419

Coverting a jpeg file to 16bit in ACR is akin to trying to make a silk purse from a sow's ear is it not? Or am I missing somethin here?

I like ACR because of some of the editing controls available... bit this will never work as well on a jpeg file... at least that is what I thought.

Just like Graystar states, 16 bit helps with the follow on math. So if you are handed a 8 bit jpeg from the camera it still helps to convert it to 16 so that you won't continue to truncate/round off to 256 levels when you do non-linear operations such as contrast/curves. When you do an operation such as contrast even on 8-bit 256 level data you generate a lot of in between the 8-bit values so it is nice to preserve them rather than truncate them.
John

Leicaman said:

Coverting a jpeg file to 16bit in ACR is akin to trying to make a
silk purse from a sow's ear is it not? Or am I missing somethin here?

Click to expand...

You're missing something. You'd convert the JPEG in the image editing software, not raw processing software.

Not everyone is editing photo as a profession or for competition. Some vacationer with a P&S that just got back from some far away land can put their pics into Photoshop Elements 6, convert to 16 bit, then use the PSE tools such as straightening horizons and building and such, and get a better result than if they just did it all in 16 bit.

Graystar I did read the post you mentioned and it is a good way to state it. In fact I read all 5 pages of posts. Very interesting thread. Nice PP job with the airplane picture BTW.

I think I asked you or Martin a question in that thread. Something about whether the exposure compensation was really helping me not blow out highlights or not.

John

The latest version of Adobe camera RAW works with TIFFs and JPGs and does have image editing ability.

John Prichard said:

Graystar I did read the post you mentioned and it is a good way to
state it. In fact I read all 5 pages of posts. Very interesting
thread. Nice PP job with the airplane picture BTW.

Click to expand...

It was a surprise when I opened the .DNG in PSE5 and it just popped up looking nearly exactly like the ISO3200 capture. PSE5 just figured out it needed +4 EC to look right. All I really did was save it!

John Prichard said:

I think I asked you or Martin a question in that thread. Something
about whether the exposure compensation was really helping me not
blow out highlights or not.

Click to expand...

My experience with the Exposure Compensation adjustment is that it will blow out highlights exactly as if you had shot at the higher ISO. However, the RAW conversion software should also have some sort of highlight compression feature which should prevent the blowing out for a couple of stops. Raw Therapee has that and it made my ISO100 +2 adjusted image much better than the image shot at ISO400. I was able to keep a good deal of highlight information. For P&S work (like what I do) it's not terribly valuable because of the noise increase. But for a DSLR it seems like an easy cheat for expanding your DR.

Leicaman said:

The latest version of Adobe camera RAW works with TIFFs and JPGs and
does have image editing ability.

Click to expand...

Yeah, I know. The RAW processor that I use, Raw Therapee, is the same way. I don't know ACR...does it have those horizon-straightening and making buildings stand straight features?

My gist of my comment is that people who aren't into heavy post-processing can use some of these canned "wizard" type image fixes built into PSE on their pics, and that by simply selecting 16 BPP from a menu before processing they can get a better result. That's all.