I’ve been playing around with RAW converters recently, and have discovered a few things.
I bought Capture One LE a while ago, and did quite a few conversions in that. Lately I’ve been playing around with Aperture and Adobe Lightroom. I discovered today that I should have done a little more homework before I bought C1LE. There’s a great little Mac-only program called “Raw Developer”:http://www.iridientdigital.com that kicks the crap out of C1LE in terms of bang for the buck. You get a pretty good RAW converter, with the full range of adjustments, some fancy curve editing and color space controls, and full-resolution fast previews (it actually seems to be the fastest or tied for the fastest in terms of 100% crop previews when compared to Aperture, Lightroom, Bibble, or SilkyPix), and it feels like a real mac program.
But programs aside, I thought I understood the RAW vs in-camera JPEG tradeoff all this time. My reasoning was that RAW was as close as to a negative you could get, and that doing post-processing in software should always get you a better result then what you could do in camera. I mean a real CPU chugging at it for 10-20 seconds should be better than what the in camera circuit can do in a fraction of a second. If I have the time to make all the adjustments, then it seems I can’t lose, right? Noise processing and resolution should both be better right?
Well the thing I learned today, was that while noise and resolution might be a tad better, color can be a whole different story. I first became curious after reading “Ken Rockwell’s infamous opinion”:http://www.kenrockwell.com/tech/raw.htm. He claims that images converted from RAW never look the same as the JPEG that the camera outputs. That sorta doesn’t make sense. I mean, given that all the data is there, any program should be able to recreate the results right? Don’t NEF’s even embed the color interpretation info right into the file itself? What kind of ambiguity could there be?
Well, I decided to actuall do a test. Here are four shots (all images were cropped manually, so they may not line up exactly. Also there is some scaling going on.. the point is to illustrate the color differences. The sharpness/noise characteristics for each image are all different — as to be expected from different RAW converters):
In camera JPEG
RAW to JPEG conversion by Lightroom
RAW to JPEG conversion by Raw Developer with default settings
RAW to JPEG conversion by Raw Developer with +42 Saturation and -3 Hue
I tried to recreate the in camera JPEG’s colors here.
The latter three shots are all based on the same NEF. The first JPEG shot is a different shot, but taken under the exact same conditions. As you can see, without significant manipulation to the conversion parameters, it was really hard to get something close to what the camera gave me. To control for the possibility of the camera actually taking two different shots between the NEF and the JPEG, I actually looked at both on the camera’s build in LCD, and the colors looks almost undetectably similar. And yet, in these JPEGs, they look enormously different.
Now which one’s more accurate? Well any photographer knows that the camera doesn’t always see what your eyes see, but after consulting with Qian (who was cutting the pictured salmon for a good 30 minutes), she agreed that the first image (the in-camera JPEG) was closer to the color of the salmon that she remembered.
So what does that mean? I guess what it means is that even though the NEF file has all the info, most RAW converter’s just don’t do a complete job of re-creating Nikon’s algorithm. That makes sense, since Nikon isn’t probably just going to tell you how they do it. But more importantly, it could be the case that using NEFs means you don’t get the colors that the camera intended to communicate. That is the part that is somewhat disconcerting to me. If my RAW converter doesn’t interpret the files (more specifically, the colors) the same way that the camera does, then how can I trust that the output of the converter is anything close to what I’m supposed to be seeing?
The one thing I did not try is to use Nikon’s own software for doing the conversion. In theory, if they manage their algorithms right, you should get similar results between their in camera jpeg and the software converter. But my Nikon capture demo license ran out. I’ll keep my files around though, in case I ever end up trying Capture NX or some other future Nikon software.
Also, If I’m really feeling up to it some day, I’ll try the same thing with an actual NEF + JPG shot from the camera, so I can really do an apples to apples comparison.