This image is from another old Ektachrome slide (34 years) sent by a friend. It also has a strong red cast overall. Faded slides are typically red because the Cyan dye has faded. Since this was a PhotoCD scan, it has to be adjusted now after the scan instead of during the scan.
This image is a perfect example because of the white fence. White is a
good neutral color without color cast. It is not white now however
(strong pink), so the neutral color allows us to easily measure the red
color cast we want to remove. The original image had some slight touchup
to remove some spots before the color correction was started.
This image project is in two parts. The first part is not so much for the procedure exactly, but instead is for background, and the desire to show you something new too. The final part will return to the method on the previous page.
Both the Curve tool and the Histogram are quite useful, and I do not mean to favor one over the other, or to suggest that one is better than the other. I will suggest that beginners normally will do better with the histogram, but as you gain experience, then you will find that you naturally gravitate to the Curve tool.
To determine the color cast situation, you can use your Eye Dropper Tool to measure the neutral tones, the white fence above is perfect for this. Pure neutral colors (white, black, gray) have equal RGB components, so any unnatural color cast can be measured. Bright white is (255, 255, 255), middle gray is (127,127,127), and black is (0,0,0). All three components are equal if neutral without color cast. But they are not equal in this case, the fence is strong pink.
In this case, the excess red cast is stronger in degree in the shadows of the fence in the lower right corner. The measurements were Red 199, Green 110, Blue 99, where G and B are less than red by about half of the red reading. But in the bright white fence in upper left corner, we see R 255, G 235, B 211, with G and B deficient about 10% or 20% of the Red reading. Blue is slightly more deficient than Green in both cases. So using constant correction everywhere cannot be perfect. The correction must also vary with intensity. It takes some experimentation. This exercise is to scope out the details that we can learn, to have a better clue.
The Curve Tool is extremely good for this because the needed correction is not linear. Several programs have a Curve tool, it is a standard tool. This is Photoshop, and most scanner twain drivers will furnish one too.
Removing Red makes the image darker, and we run out of range too. So we add Blue and Green instead. To do that in this case, I boosted both of the Green and Blue channels of the Curve Tool as shown above in the Green channel. I left the Red channel alone. This is better done in the scanner during the scan, because the 30 and 36 bit scanners have more range internally before saturation occurs than does the 24 bit output image that we have now. But this image was already an existing PhotoCD scan.
The Curve Tool description explains how the Curve Tool is the scanner's response function, and that the bottom scale is the input, and the left scale is the output. So compared to the straight 45 degree default line from corner to corner, the values in this setting are instead mapped this way (see chart) to add greater boost to Blue and Green in the dark tones than in the bright tones.
| Input | Output | Increase |
| 100% | 100% | 0% |
| 75% | 100% | 33% |
| 50% | 100% | 100% |
| 25% | 50% | 100% |
| 1% | 2% | 100% |
Both the Green and Blue channel have values between 50% to 100% luminance changed to output at 100% (as shown, blue is perhaps 45% to 100%, slightly deeper than green). This brightens Green and Blue, per the chart above, offsetting the excessive Red. The Red channel was not adjusted.
In the dark range 0% to 50%, the Green and Blue curves are steepened considerably. We are brightening the Green and Blue shadows much more than the highlights
This curve is the identical action as moving the histogram White Point to the mid point. Both of those actions are the same, and hopefully that relationship is becoming clear. The main difference is that the Curve tool will do many more things, and the histogram does fewer things
I used a "straight line" mode above instead of the "curved" curve more normally used. That is not better or more accurate, it is hardly different from the curve shown at right from the last chapter. But in this case, it was easier for the preparation of the chart, and perhaps is easier to understand the technique of the input/output tone conversion. Above, I simply grabbed the White end point with the mouse, and slid it over to the midpoint, in both the Blue and Green channels, but not the red channel. You can use the curve tool to modify the response as you wish.
Given that the Green channel is already adjusted as shown above, now we
are looking a green image. We still need to adjust the Blue channel in
the same way to balance the color. In this case, the yellow is because
not enough blue is added. The Blue White Point is moved to the 75% point.
Yellow is opposite Blue on the color wheel.
Too much blue is added. The Blue White Point is moved to the 25% point.
Goldilocks Blue, just right, the Blue White Point is moved to the 50%
point. We added a little more Blue than Green (shown above). We set the
Green channel setting based on the measurements and our logic. Then we
set the Blue channel by eye to balance it, and it came out about as
expected. We have doubled the Green and Blue values in the dark half (see
chart), and added much less than that in the highlights.
This procedure did give an image with good color balance. This method is just one of several possible, and is not at all a hard rule. The Curve Tool and Histogram Tools will be your best tools for this type of work, because they allow more control than the simpler tools. You often can not use the same treatment in both the shadows and highlights, which is all the simple tools normally do.
Then on next page, we start over using the histogram instead. The above is possibly better for understanding, but the histogram is possibly more convenient.
Copyright © 1997-2010 by Wayne Fulton - All rights are reserved.
