K10D, WB and color histograms

Started Apr 23, 2007 | Discussions thread
OP Jonas B Forum Pro • Posts: 14,596
Re: K10D, WB and color histograms

Part I

Background:
===========

This is a follow up to Andrews thread here:
http://forums.dpreview.com/forums/read.asp?forum=1036&message=22765999

A bit down in the thread we discussed histograms and what they tell us. I described an approach to "calibrate" my 5D's color histogram with the reality (as seen by a raw converter) I did. Interesting to some, boring and an irrelevant waste of time to others. Now that I have the K10D I have done the same thing, at least a first round (and there it probably stops).

When "exposing to the right" we trust the histogram and the incamera preview blinking to ensure there is no overexposure going on. Or rather, we try to keep the overexposure reasonably limited to spectacular highlights which we don't care for, exposure wise.

Now we really shouldn't trust the histogram or the blinkies. They show what's going on, but roughly, only. To be able to make the best possible out of it we have learn the tools.

Purpose:
========

  • Find camera settings that make the histograms more reliable.

General disclaimer:
===================

I have had the K10D in my hands for less than a week. My understanding of this sort of things is limited and my conclusions may be off. I'll be happy to stand corrected.

Method:
=======

I have used the K10D, a greycard, a Gretag CC chart, ACR4 and the Malykh PEF analyzer togehter with a couple of bulbs ("daylight type") and typical Swedish afternoon overcast. As you can see I have taken science to a new and high level here.

Kelvin
======

Anyway, first I wanted to see at what color temperature setting I can get a color histogram to be reasonably close to the degree of saturation at the sensor.

I used the same method as I used with the 5D: I took a long series of pictures of the grey card with the daylight bulbs as light source. By stepping the kelvin setting with K200 per picture and overexposing 1 stop I hoped to find a setting where the blue and red channels in the histogram and in relation to eachother showed the same result as the output from the Malykh analyzer.

Clicking away, analyzing, comparing... 15 minutes of work showed that K4800-K5000 is right. (Trivia: This is a bit lower than the 5D where K5500 is a good setting.)

That was the first and easy part. I now have the camera set to K5000 all the time.

Why Kelvin?
===========

Imagine you are taking pictures at home, the light source may be a mix of halogen lamps and (probably mainly) ordinary bulbs. Now you set the White Balance control on the camera to Bulbs. Take a picture of a bulb and it is showed as a white bulb on the LCD. Not good? Now look at the color histogram. All three channels are at the same level and there probably is some headroom. Not good? No, this is not good if you try to expose to the right. Try it; dial in +1.5 as exposure compensation and take that picture once again. Everything is still fine. When opening the picture in ACR (or any other raw converter) you'll find that the red channel is blown away. By using another setting we can obviously get a better grip of the channels.

Light source; Bulbs, daylight types
===================================

At the border of blowing it
===========================

The next thing to do was to exchange the grecard to the Gretag standard 24 patches Color Chart. The light was the same. I took a long row of pictures, now with the WB set to K5000 all the time, but different EV values, in steps of 1/3 stop. I started way down and clicked away until I was way overexposing the pictures. Analyzing and comparing followed.

Now I found differences between the incamera histogram, the blinkies and the Malykh tool.

At EV + 1 and 2/3 we get this picture:

We can see the color shifting in the bottom row and the white patch is blown.

The corresponding Malyk analyzer result picture is here:

As we can see there no channel clipping when looking at the raw data from the sensor.

The first picture above is the result from ACR 4 with all settings to neutral, straight, plain, untouched and zero respectively. This means default values except for the curve that was set to linear, and sharpening and noise reduction which both were set to zero.

When adjusting the WB in ACR things got better:

Now it is only the yellow patch that is blown.

Adjusting the EV value in ACR takes care of this:

The picture is still somewhat overexposed, paired with a low contrast. We'll have problems keeping the yellow at place if we increase the contrast.

===================================
to be continued
===================================

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