Sunday, November 27, 2005

The Camera Profiling Experiment Rules

I've started to attract a lot of visitors interested by my offer of free camera profiles. I'm not doing this out of the goodness of my heart, I'm doing it to gather experience in what works and what doesn't.

You send me a target, and a couple of photos created with the same lighting and settings. Preferably all of the above are in-camera Jpegs, otherwise identical Raw conversions. No photos that allow me to evaluate the work means no profiles.

If you send me in-camera Jpegs, that means 2 files (One chart, one sample).

If you send me conversions, that means 4 files, preferably:(Converted chart, converted sample, Raw chart, Raw sample).

Large files can be easily sent via

I'm summarizing the shooting guidelines below:

The target has to be very carefully lit, any gradient or colored reflection will damage the profile. One single softbox far away is good lighting, it can be masked or goboed to even it out. A mini colorchecker may have less lighting issues than the full-size one.

The target should not occupy the whole frame, to avoid vignetting prolems leading, again, to non-uniform lighting. It should not be cropped, I want to see the cardboard frame at the very least. If possible it should be mounted on some grey background.

Yes, lighting is crucial. The method I know off I learnt from the Coloreyes software - a respected profiling tool - which recommends ONE single softbox, slightly to one side, as far away from the chart as possible, feathered or goboed to make the illumination constant. An additional precaution is to move the card holder away from the back wall, in fact keep it away from every wall because of reflections, and also pay great attention to reflections on the camera. One thing I might do in the future is recommend two shots of a rotated target and create software to help me average out those shots.

TESTING: "The camera should be custom gray-balanced using whatever tools you usually employ for this. You can gray-balance in the Raw converter by clicking on your preferred target square."

The workflow must be locked down. Fixed conversionparameters, or in-camera Jpegs, please. The camera should be custom grey-balanced. And, if possible, the Lab value of the colorchecker white patch should be around 96, if this is the target you're using. You should almost be blowing out that patch, I think. Also, beware of lens vignetting ...

Canon's DPP is a very good converter to attempt this with.

When you get the profile emailed back, you simply assign it to your images in Photoshop.

You can send files to the address at the top of the blog. However, I will expect feedback as a condition of making the profile. A couple of additional images taken with the camera in similar lighting, sent with the target file are one of the requested peices of feedback.


No comments :