The Mystery of RAW (Discussion)

I stayed up until 3:12AM for this damn article.

New Article: The Mystery of RAW

Please tell me if I am totally wrong in my conclusions.

Eric fascinating to see the difference in the RAW converter software and the sunbursts. The dolphin example is very clear. There is a massive difference between the images. And there was no way you could replicate the Canon RAW conversion using the Adobe software?

Thank you for sharing this with us. Invaluable stats! Your article is churning out of my printer as we speak!

Have you (or has anyone else) tried the either the Nikon or Fuji supplied converters to see if they have the same advantages over Adobe that the Canon one has when it come to sunbursts?

Brilliant!

Alex

Hi Eric,

I read your article with high interest and had to laugh like hell in some passages where you mentioned the JPEG-conflict between professionals like Jim Watt and young digital rebels like us.

I theoretically agree, but was not convinced with some sample posts you made to underline your conclusions. For example I like the crop of the "Adobe Photoshop CS - Shadow: 5" best. Also I see an increase in the performance of Capture One and your article makes me want to upgrade to v 3.5. The photos processed with Canon software seemed too bright for me, actually containing more white than the others. Did you probably mix up the image comments?

another reason might be that I am currently surfin via a 30" diameter TFT display I call now my own (could not resist a local special offer). I expect the pictures to be generally brighter than on my calibratet CRT Monitor. I will check them again later with it and inform you about my opinion.

Another thing that surprised me, was that you did not present results converted with Breezebrowser. Chris Breeze also claims to have an optimized algorythm and I'd love to see it next to the others.

That's all for now...

I have to start packing and get some sleep as I am heading off to Sweden tomorrow.

Andi

Here are two sunbursts shot on the Fuji S2 in raw mode with auto exposure using the 10.5mm Nikon Fisheye in Aquatica housing.

First one converted with Photoshop CS, auto WB, 0 settings but contrast 50.

Second conversion with the Fuji EX converter.

Settings as camera - Org Colour, Org Tone, Sharpening off.

Andi -

I'll have to add Chris' converter to the mix. I had assumed that he was using the Canon SDK without optimization, but I may be wrong.

Andi - I did not mix up image comments. The Canon does show more white than the others do, specifically (as I mentioned) in areas that are very close to white. I perceive the gradient to be smoother, whereas the Photoshop and CaptureOne conversions jump from turquoise to white in one step. Hence, the Canon ones must be brighter.

Also interesting is that there is no way I can get rid of the jaggy white, even using curves in CaptureOne. And at first glance, C1 3.5 doesn't do any better than C1 1.2. The v3.5 conversions I added were underexposed until the red overexposure warning was turned off (as mentioned in the caption), so they definitely look different than the default curve setting in the C1 1.2 conversions. You should NOT conclude from the article that C1 3.5 does a better job! (There isn't enough info).

I'll add to the article eventually:

- BreezeBrowser conversion

- Yarc conversion

- Notes about C1 3.5 vs C1 1.2

- Notes about CA correction alternatives to Photoshop CS converter

And if someone can get me a suite of images converted in a similar manner, in the Nikon/Fuji camps, I can update the article with those conversions. Please don't do it on the forums -- get all the images ready, stick it in a zip file, and then get in touch with me.

Thank you for sharing this with us. Invaluable stats! Your article is churning out of my printer as we speak!

I hope you are using the printable feature of the site? You know, that little printer icon next to the feature?

Eric, most interesting article. I know this discussion relates to how different Raw converters handle sunballs. But I've been really fascinated by them as a focus point (Maybe to an excess). Keep us updated on this project. If you find the software that does the job, I'd be first in line.

In the meantime, I'll continue on my quest to capture the perfect "Sunball" shot. (Wonder if it will have a model in it?). :roll:

Rand

In the meantime, I'll continue on my quest to capture the perfect "Sunball" shot. (Wonder if it will have a model in it?). :roll:  

 

Rand

Rand - I think the perfect sunball shot may have film in it, and then maybe a model.

Thank's for the enlightening article.

There's no doubt that starting from a raw file and custom tweeking the curves is the best way to display the details in both brightness extreems in a photo with a large dynamic range. However, I think the availbilty of more details in the highlights in the raw file depends on the camera vendor. Having recently switched from the Olympus 5050 to the Canon DRebel, I was surprised by how much more details I can recover from the Canon CRW file comparied to the Olympus ORF. The other thing I noticed about the Canon is that I found clipped highlight in the jpeg part of the CRW file when I thought I had properly "exposed to the right" based on the histogram, highlights which I later recovered in the raw conversion process. In short, I think Canon clips the highlights in the JPEG conversion much more so than Olympus.

So far, I've tried C1, Canon EosViewer, and dcraw. I think the best results are from C1, followed by dcraw + photoshop, and the EosViewer is the worst. This may partly be due to the better interface of C1 where I can do more tweeking and get better feedback before converting.

Is capture 1 only available to Cannon camera owners? By that I mean you must purchase the camera.

Herb -

With the dolphin file, I am unable to get ANY good results out of C1, even if I tweak the curves. When the burst area starts to look ... slightly acceptable, the rest of the image is so dark that you'd have to extract a second image out of it and composite. Canon EOS Viewer and Digital Photo Professional do a better job of getting out a single image that is acceptable, for that image, anyway.

Perhaps this could be rectified by a custom colorspace profile.

Note that in general I prefer C1 over the others, but for sunballs, C1 is horrible.

Eric,

I've only been messing with CRW files for a couple of weeks and I don't have some of the converters you do, but I like the performance of the DRebal and C1 on Sunball shots. I'm not as prolific at this type of shot as Rand, but I did try quite a few on my Thailand trip, and these are the type shots that I've used to compare the 3 converters that I have on hand. I think C1 is better.

In addition to the ones I posted earlier, here are a few more where I think the sun came out well.

Those sunballs aren't bad, Herb.

Maybe it's the 1Ds that is bad with C1? Can I get those RAW files from you to some tests myself? I won't distribute.

Maybe it's the 1Ds that is bad with C1?  Can I get those RAW files from you to some tests myself?  I won't distribute.

I'd love to see what you can do with them. I don't have a server that can host the 7MB files. Do you have a place where I can send them via ftp?

Eric,

I've noticed the exact same posterization in shadow areas with ACR, in fact this all started with the 2.2 version that added support for the Mark II.

I posted about this issue a few weeks ago in both Rob Galbraith and DPreview. Some people have chimed in to agree that ACR exacerbates posterization. No one has produced similar results though, which is why I was glad to read your article.

Here is the link to my DPreview post, with sample images & shadow posterization at ISO 250! It's not only high ISOs, I've gotten it at 250 and 400 too.

http://forums.dpreview.com/forums/read.asp...message=8996064

Anyhow, I have my suspicions that it might also be a MarkII related thing. My 10D files also gets posterized with similar settings, but not as badly. Many people have posted MarkII files converted with ACR at various shadow slider settings and not gotten any posterization.

I've ordered another MarkII to see if this hypothesis is correct.

Yeang

Herb and Eric,

What shutter speed are you shooting those sunballs at?

On many occasions, my 10D produces sunballs with smooth gradients like what Eric's shown, using EVU or FVU to convert. So it's not solely a 1Ds thing.

Yeang

I've ordered another MarkII to see if this hypothesis is correct.

Could you send me one for independent confirmation

I shoot most of the sunball pics at 1/320. Yes. The DRebel will sync at that speed even though it's spec at 1/200.

With the dolphin file, I am unable to get ANY good results out of C1, even if I tweak the curves.  When the burst area starts to look ... slightly acceptable, the rest of the image is so dark that you'd have to extract a second image out of it and composite.

Hi Eric

An excellent article. Just a quick question. In the quote above you talk about compositing the shot. Have you tried the shadows and highlights filter in photoshop? I have used it a couple of times and it seems to work quite well. I was wondering if you could adjust the RAW image to balance the highlights and low lights as well as you can and then try messing around with them inside photoshop using the highligh/lowlight adjustment.

Richard