Trevor G: Now I understand...

Started Jan 12, 2013 | Discussions thread
Mark H
Senior MemberPosts: 2,871
Like?
Re: ...Highlight Recovery. RAW can do what the JPEG can't [still wrong]
In reply to Trevor G, Jan 22, 2013

Trevor G wrote:

Mark H wrote:

Trevor G wrote:

The real reason that EXR was introduced was to increase dynamic range.

Yes - but, also to more efficiently reduce noise (particularly at high ISO) via 'EXR-SN'.

This is achieved by allowing the sensor to receive a higher exposure level without clipping the data.

That's incorrect - the sensor reduces clipping by using the same and/or a lower exposure.

This occurs in M size when DR is greater than ISO, and is sensor-based, in that adjacent pairs of pixels are exposed differently and the resulting 2 images are combined to give an extra 2EV of highlight recovery or headroom.

Not really "recovery" as such - it 'extends' the range of the highlights captured (in both JPEG and RAW).

Mark, it's nice of you to drop in and bless us with your expertise.

Silly and ill-judged patronisation isn't going to get you anywhere here.

I'm unable to respond fully at the moment, but I have posted plenty of images with histograms which show these points. Many of these were in November 2011. You'll find them as original threads of mine.

I've seen you previous post before (not necessarily every single one) - but they do not/cannot support your claims and assertions, because they are clearly wrong, and very easily disproved.

These extra 2EV of highlights are only available when processing RAW - the in-cam JPEG clips away the upper 1EV to 1.3EV.

That is completely wrong.

See above.

I would say the histograms prove my points.

This is where you are wrong/mistaken - the histograms don't prove anything of the kind.

In general, histograms are not EV linear (in terms of 'exposure value') and you cannot use them to measure exposure range.

Virtually all JPEG (or processing RAW) images use non-linear tone curves, and Fuji's extended +DR% tone curves are even less linear/more exaggeratedly curved in their highlights than normal.

The mistake I believe you are making here, is that you are failing to appreciate (as I've had to explain to some other's on these forum before) that the extra +2EV of highlight range in a DR400% JPEG is actually compressed (contrast reduced) into the same space as only 0.6EV of highlights occupy in a regular DR100% JPEG...



Note: the '91.6%' above should read '90.4%' (not that it matters)

...and quite clearly, the above test graph shows that the DR400% JPEG contains all the extended +2EV highlight range - it is not clipped off at all - and it is relatively quick and easy to bring that +2EV out better in the image by curve adjustments such as PS 'Shadows/Highlights' adjustment should anyone wish to do so, just as I have done with your OOC JPEG example.

I notice you didn't post a screen shot of the histogram to attach to your image...

And neither did you post any histogram on your original message that I replied to.

But regardless, as I said before, histograms are actually of little or no use in showing actual absolute exposure values/range.

- I'll do it for you:

Let's compare it with the histogram on my RAW:


My eyes see quite a difference in the highlight end of the histogram.

There is no 'significant' difference between those histograms.

The little difference that there is at the highlight end, is simply that my original PS 'highlight' adjustment has resulted in a slight lowering in the amount of maximum white, something that is instantly rectified by a slight tweak of 'levels'.

There is always bound to be differences between these histograms, simply because we have used two quite different adjustment routes, and two different software applications, to produce 'nominally' the same effect.

It's hard to see the effects of that compression on your intzy wintzy little spider image...

If it pleases you, here is a full size copy of the same again (albeit this time I've made a slight tweak to the white point level as I mention above)...

1653045 (6MP) OOC JPEG 'PS Highlights +50%' adjusted

There are certainly slight qualitative differences pros/cons, between my PS adjusted camera JPEG and your Silkypixs RAW version...

The PS highlights adjusted JPEG has rougher highlights due to the stretching out of the limited number of original highlight tones (and no doubt compounded by initial JPEG compression).

Conversely, the Silkypixs RAW version has significantly worse noise levels in the lower tones, not to mention deviating significantly in colour tones from the original camera JPEG output colour.

But, all that is entirely irrelevant to the original point of contention here, i.e. you original claim that the camera JPEG  "clips away the upper 1EV to 1.3EV" and that the full extra highlights range/detail etc is "...only available when processing RAW" - something which both DPReview's test chart, and my adjustment of the OOC JPEG clearly disprove.

It's preposterous to suggest that the camera captures +2EV, but then only uses as little as +0.7EV to 1EV of that extra +2EV in it's JPEG - you simply aren't measuring/interpreting the data correctly.

Here's an example - the OOC JPEG:

If you play with curves you will get some more highlight detail back, but that will also change the colour tone.

Here is the matching RAW [image bottom of page] , after having its exposure dropped by 1.5EV to bring the highlights back from clipping, and then having the lowlights tweaked (raised) to give the image the same overall brightness as the JPEG:

It's not a matter of any 'clipping' at all (not even in the OOC JPEG).

Mark, the images I showed are how they appear as JPEGs. If you have Photoshop (I don't) and can pull back highlight data from the JPEG then that is great!

It's not a matter of being "great".

The purpose here is simply to show you that your original claims and assertions about the JPEG highlight range were patently wrong/mistaken.

However, as it stands, using Any of the JPEG software I have on hand, I cannot do that.

I dare say - but that doesn't actually justify your repeatedly making the incorrect and misleading claims and assertions as you have.

The good feature abouyt using RAW is that you simply reduce exposure in software like Silkypix and the over-exposed highlights come back down into view. Applying the same simple technique to JPEGs just does not work.

I would suggest that you try to get some better software - there are easier ways of doing it.

You don't need the 'full' Photoshop either - both 'Photoshop Elements' and 'Lightroom' have the same 'Shadows/Highlights' adjustment capability (to varying degrees of complexity).

As I said before, if you can do that in Photoshop, that's great! I'll give you some images with greater over-exposure but which will fully unclip from the RAW, without loss of colour tone fidelity.

Other than about a fractional +0.2EV of RAW headroom (as I've carefully measured it, here ), neither RAW nor JPEG can recover real 'over-exposure', i.e. 'clipped data'.

My purpose in this thread/sub-thread is not to claim that JPEG is a match for RAW in all circumstances - it was only to debunk your false/incorrect claim that Fuji's JPEGs somehow "...clips away the upper 1EV to 1.3EV" of it's DR400% - which is/was thoroughly misguided and misleading.

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow