This is my no means scientific. I simply processed a shot through 4 different RAW converters (WB tweaked by clicking on the shirt, levels tweaked, hilights tweaked where needed - all basic stuff done in less than 2 minutes). I purposely chose a 50 1.4 shot wide open in natural light (indoors with window light) at ISO 1600 since I was most curious about how each converted rendered noise (no noise reduction applied in converting) and WB.

The software compare:
Adobe Camera Raw 3.1
Canon Digital Photo Professional 1.6.1
Bibble 4.2.6
Raw Shooter Essential 1.1.2

Here's the 4-way comparison:

Great comparison, thanks I find the ACR noise reduction pretty crappy, I turn it off and let noise ninja do it all inside my batch convert action.

Edited by Tim Wild on May 18, 2005 at 11:48 AM GMT

Tim Wild wrote:
Great comparison, thanks I find the ACR noise reduction pretty crappy, I turn it off and let noise ninja do it all inside my batch convert action.

My pleasure Tim. Note that I didn't use any noise reduction - I wanted to see how the converters RENDERED noise and ACR rendered the least color noise of the bunch.

I think the defaults of ACR do a little sharpening and noise reduction, but it depends how you have yours set up, or you could have turned it off.

Tim Wild wrote:
I think the defaults of ACR do a little sharpening and noise reduction, but it depends how you have yours set up, or you could have turned it off.

I do turn off NR and sharpening in ACR - I like to control both aspects in PS.

Thanks,

Joe

Thanks, Joe, this is a good simple comparision. Although I have not used RAW Shooter, I tend to agree that the ACR is the best unprocessed beyond where you left it, and I agree that the sharpening and other stuff is best done in PS/CS2. Thanks for posting this.

What would these files look like without all your tweeking and just convert them "as shot"?
Steve

Steve Tinetti wrote:
What would these files look like without all your tweeking and just convert them "as shot"?
Steve

I'd be interested in this as well, just to eliminate any unintentional 'human' interference.

Just keep in mind, that "as shot" mean different things to different converters. To DPP, it's the parameter settings on the camera. For RSE, camera settings are ignored, and the "default" is tweakable on a per camera, and per ISO basis. In looking at the "test," it seems fairly clear to me that ACR is doing some color noise reduction--turning off NR there probably simply turns off luminance noise reduction.

your tests demonstrate nicely the problem with skin shadows I have always been having with ACR. There always seems to be very little colour depth in these areas, everything resolves to 3 or 4 tones of browns or yellows. Your RSE example shows better shadow resolution, but overall the portrait is much flatter than the ACR example - something I've been seeing a lot of in my results. Otherwise I'm reasaonably impressed with the Pixmantic product.

Great test. When I first looked, the ACR was the brightest and appeared to be the "best". Now, I'm not so sure.

The picture is flatter than the DPP rendering. Only you can tell us which skin/hair colors are closer to your subject's, but the skin tones of DPP look more real (albeit with somewhat of a gray cast that could be removed). Look at the right eye (on the left for those who don't know right from left ). There is clearly more depth in the DPP version.

Jim

Would be interested to see a result from EVU. My own experience shows that EVU gives the best skin tone for RAW conversion.

Perry

great test. i'd be interested to see how it handles different types of subjects, like landscapes, and at differnet iso's.

the ACR looks best to me in the images your provided. thanks for sharing!

emo

Guy Mancuso wrote:
Be nice to have C1 in there but what i find weird is the colors look better or more natural in RSE but that is were they are the weakest otherwise I would switch to them from C1. I will admit been palying around with the new Arc 3.1 and it is much better at workflow but still can use improvement there but the images are looking better
The seem to have been working on colour since V1.0 (which was somewhat strange in other than flat mode).

"...but what i find weird is the colors look better or more natural in RSE..."

I think what you're seeing is the blue cast that has been added.

Please forgive me if you find my following to be nutty thinking.

Put some blue glasses on and go outside, high noon, blue skies with a few fluffy clouds in the sky. Where the glasses for a few minutes. Raise the glasses, lower the glasses and then raise them again.

In my case, I became painfully aware of how much yellow/jaundice is in reflected sunlight as opposed to the blue cast we've all been taught that is natural daylight.

It ain't the way we've been taught to see folks.

Look at the clouds in the distance and you'll see the clouds, if the sun is out, have a bit of a yellowish/jaundice look to them as opposed to being white or grayish white as we've been taught to think of the coloration of clouds.

We are clouding (no pun intended) our seeing with what we've been taught as opposed to seeing things as they really are.

Please don't lock me up for what I wrote above.

I'd love to see you do this test again, but at least match the exposure/contrast levels. I don't think any of use the defaults these tools give us.

I might be missing something, but to me the Canon image processing looks more realistic. The others are nice and warm. I agree they look nicer, but I think the Canon looks more realistic.??

For me it's all about workflow. I know that C1 and RSE can produce results close to DPP, but I've found that DPP gets me where I want to be much more quickly - basically I end up just spending a long time tweaking C1 and RSE to give me what DPP would give me by default. Bibble is a total no-go for me since it simply can't deal with super-saturated color channels in my concert photography work.

I agree with Guy, a color chart would be a good idea. Perhaps even a custom white balance shot. Workflow is indeed important also. Since many images still need to some PS tweeking anyway, a converter that allows you the best combination of quality and speed might be best.

With all due respect to the original poster, who specifically stated this was not scientific; this type of "test" should be conducted under more controlled conditions, to narrow the subjective to the objective, IMO. As it is, we're going back and forth based on our subjective opinions of a subjective test shot under subjective conditions. Without a known value anywhere in the shot, how can we agree on anything?
Steve

Steve Tinetti wrote:
I agree with Guy, a color chart would be a good idea. Perhaps even a custom white balance shot. Workflow is indeed important also. Since many images still need to some PS tweeking anyway, a converter that allows you the best combination of quality and speed might be best.

With all due respect to the original poster, who specifically stated this was not scientific; this type of "test" should be conducted under more controlled conditions, to narrow the subjective to the objective, IMO. As it is, we're going back and forth based on our subjective opinions of a subjective test shot under subjective conditions. Without a known value anywhere in the shot, how can we agree on anything?
Steve

Hi Steve,

Thanks for the respect - this response is not a disagreement, it's just my rationale for posting. I actually chose my subject title well and called this a comparison. As a test it's fairly useless - as a comparison it has some value. I approach RAW conversion from a subjective perspective - I have a "look" I want to achieve based on how I saw the shot when captured and a "time value" relating to the effort it takes to achieve that look. I showed this comparison of my basic "quick conversion" using as many pre-sets as possible to get a good result. I subjectively find ACR the best output but RSE the easiest to achieve "close enough".

I do like your idea of a controlled test and I will do so with a GM color chart because I too am curious how each converter handles WB by clicking a color patch. It will also provide objectivity for a entirely different evaluation of RAW converters.

Thanks for posting.

Ranger099 wrote:
My pleasure Tim. Note that I didn't use any noise reduction - I wanted to see how the converters RENDERED noise and ACR rendered the least color noise of the bunch.


Are you sure you weren't using the default settings in ACR, which is +25 for Color Noise? If this is a Canon CMOS sensor, I doubt that CNR wasn't on, looking at the ACR example...

Jeff wrote:
Ranger099 wrote:
My pleasure Tim. Note that I didn't use any noise reduction - I wanted to see how the converters RENDERED noise and ACR rendered the least color noise of the bunch.


Are you sure you weren't using the default settings in ACR, which is +25 for Color Noise? If this is a Canon CMOS sensor, I doubt that CNR wasn't on, looking at the ACR example...

You are correct. My mistake. I ALWAYS turn off CNR but missed it on this shot (I just ran the file through ACR and it's clear that my comparison result had CNR +25). Sorry for any confusion.

As I mentioned above - I'm actually going to re-do this as a more formal test with tight controls.

What curve did you use in Bibble? I find the default oversaturated look unuseable, but w/ the colorimetric curve it matches C1's and RSE's look, and becomes the best RAW converter for me to work with.

Jeff wrote:
Ranger099 wrote:
My pleasure Tim. Note that I didn't use any noise reduction - I wanted to see how the converters RENDERED noise and ACR rendered the least color noise of the bunch.


Are you sure you weren't using the default settings in ACR, which is +25 for Color Noise? If this is a Canon CMOS sensor, I doubt that CNR wasn't on, looking at the ACR example...

OK, dumb dumb question here
+25 color noise means adding color noise by 25%?

Tight controls would not include "clicking on the shirt" to set the white balance. The shirt is patterned. Even if it were not, a pixel or two over will give you a different white balance. It is impossible to get consistency that way.