Lightroom, color correction and D. Margulis

canoflancanoflan Registered Users Posts: 168 Major grins
edited June 21, 2007 in Finishing School
I am sure there are many on this forum who have learned a great deal about color correction from Dan Margulis's "Professional Photoshop, 5th Edition," like I have. I have found it indispencible for getting photos to be "believable" in color first, before doing any increases in contrast, or saturation/vibrance. For those that may read this post and not know what the technique basically is; well, you basically take readings of known colors in the info palette of PS and see if those numbers line up with what you would expect a reasonable person to believe the color really is. Additionally, Dan believes that since the most reasonable way to introduce contrast is via curves, you must decide what part of the image is most important and ensure the steepness of a luminosity curve covers that range.

That being said, I have learned how to get the most of the LR in using this technique and here is the step-by-step:


1) Upon importing into the Library of LR, don't do any adjustments except defaults. It is important at this point no changes to the original color of the original occurs.
1.5) If you have a noisy photo, anything you do in the Develop module makes it worse; therefore, use the Develop module color noise slider to remove any color noise, first, then go into Photoshop and remove the luminance noise with the noise filter or your favorite after market noise removal product. You may want to do a quick initial adjustment to white balance before doing this step, but the sliders are always there anyway upon returning to LR with PSD file. You may try do the following steps, then go into PS to remove noise and sharpen, but it is really up to you and the areas the noise is prominent. If noise is only in insignificant areas, then I wouldn't get concerned about this step.
2) Adjust exposure, recov, blacks and fill light (be careful with fill light) to get the photo exposed the way you want it regarding luminosity. Don't use brightness or contrast sliders. Leave them at their default settings.
3) Begin to take readings using the white balance dropper, but don't actually use the white balance dropper by clicking on areas until you perform this step of taking readings. Moving the white balance dropper around, write down (since you cannot record various readings in LR) the readings of areas where you know what the color must be, as best you can. Take readings of all areas of significance where you know what the color must be. If that is 2 readings (like white and black for a tuxedo) then fine; but if for a very colorful summer day at a flower garden, take 10 if you have to.
4) At this point, either you need the color palette in PS, or Dan's book to understand what RGB values yield types of color (i.e. green, yellowish green, greenish yellow, red, reddish yellow, etc...). Ideally, we would have LAB and CMYK value abilities, but those may be coming in a later version of LR. Using the color pallet in PS LR, you can use the other color spaces to generate the colors you expect and see in RGB how close your readings are. I know that LR uses percentages instead of 0-255, but those of you used to CMYK, it is simply 50% equals 128 lightness in RGB.
5) After understanding whether your colors are reasonably within the range of expectation, or out of expectation, you begin to use the white balance dropper to make an initial correction by clicking somewhere you think should be gray (in the 70%,70%,70% range in LR for RGB). Use the left side of the module that changes in the picture as you move the WB dropper around. As you click and the color appears better, check your color sampled areas again for accuracy and see if they are closer to the expected color or not. If closer, you should use the WB sliders to fine tune until it is right; if further away, then you need to keep on with the WB dropper until you move in the right direction, at least.
6) Now, after you have fine tuned with the WB sliders, some areas may still be out, or not enough in the direction you want to go, due to things like conflicting color casts, etc.... Some of these can only be fixed in PS with masking techniques, however, try the HSL sliders using the dropper type instrument that you can pinpoint an area, click and move up and down directly in the photo and the appropriate sliders corresponding to that color will move left with downward movement, and right with upward movement. If you are trying to change a color, you may best be served starting with Hue sliders. If you need more of the same color, use Saturation. If you need the color lighter, or darker, use Luminance. It doesn't take long to focus on an area, try H, S, or L and see if your colors get closer to what you want. Remember, H changes color, S increases or decreased that exact color in terms of very little color (therefore white) to very intense, and L moves from gray (no color) to the actual color.
7) At this point, go up to the curves in the Develop module and mouse around the significant part of the picture to see where on the curve it lies. When you have this understanding, then move the curves to ensure the significant area is in the steepest part of the curve.
8) If you know about how to intensify color through LAB and need more control over contrast, at this point, move into PS and covert to LAB color space profile and perform curves based on Dan's book for simple curving.
9) Size, Sharpen and Print.
10) Enjoy!

Let me know please if you have better suggestions for I am always seeking to learn.

Pat

Comments

  • arodneyarodney Registered Users Posts: 2,005 Major grins
    edited June 20, 2007
    canoflan wrote:

    Let me know please if you have better suggestions for I am always seeking to learn.

    Pat

    My comments would be that LR isn't Photoshop, raw files are not rendered baked images you'd work with like those in Photoshop.

    Work with the tools as in the designed order: Top down, left to right.

    Many of the tonal tools may not be needed (don't think just because there are curves, you SHOULD use them). The Paramretic curves in LR don't operate like Photoshop by design and for good reasons. They are often only necessary for very fine tune work that as I said, often are not needed if you fix the big stuff in Basic. There are two places where the tone corrections are applied and this is why the controls are split up (and why you notice the Histogram in Curves, which is based on settings applied in Basic) look different.

    The WB tool should be used on a white but not specular. So your values of 70 are way off here. This is also taking place in a linear wide gamut (ProPhoto RGB primaries) internal space, the numbers and histogram you see are not. WB should be job #1 hence its position in basic.

    You'll be waiting a very long time for LAB or CMYK values in LR.

    The idea that a noisy image will get worse if you tone correct is news to me; if anything, you want to apply the edits (in one convolution which is what a metadata editor does) in high bit linear encoded data where you haven't tried to spread small amounts of shadow data into a gamma corrected space and spread it even further in Photoshop. IOW, I'd like to see an example where you applied tone corrections in LR versus Photoshop to produce the same tonal appearance then examine the noise.

    There's really no need to complicate the way LR handles tone and color corrections. We're basically rendering Grayscale data into a color image, it should not be thought of as color correction but color creation!

    Its useful to understand the significant differences between scene referred and output referred images. Here's a good primer. Raw processing is taking scene referred to output referred processing.

    http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf
    Andrew Rodney
    Author "Color Management for Photographers"
    http://www.digitaldog.net/
  • canoflancanoflan Registered Users Posts: 168 Major grins
    edited June 20, 2007
    Therefore...
    Andrew,

    Thanks for the info and it is very thought provoking to say the least. I am taking away from your reply that doing all we can in LR is the best way to go about addressing major changes to the RAW data for WB and tonal corrections, including noise suppression.

    Forgive me if I don't appear to catch all you stated in your reply for I am a bit of a "bottom line" type that likes to boil things down to "here is what you should do to use the tool right" type of person. Therefore, I hear what you saying as use LR to the full extent possible before going to PS because of the flexibility we have with the RAW file and the control LR gives us.

    Am I on track? Let me know.
  • arodneyarodney Registered Users Posts: 2,005 Major grins
    edited June 20, 2007
    There are workflow and quality advantages to doing all the heavy lifting in LR using the linear encoded, high bit, wide gamut pipeline from Raw data.

    Photoshop is great at correcting rendered images where you basically need to change the existing baked in numbers that are represented by (often) million of individual pixels. This isn't how LR operates with raw data (using existing rendered images, its questionable if you should do anything other than use LR as a database although the jury is out). For raw, its not anything like how images are manipulated as we've done in Photoshop for 17 years.

    Both Thomas Knoll who wrote ACR and Mark Hamburg, the chief architect of LR have attempted to place the controls to best handle this raw data and its just different than working with pixels in Photoshop. Its important not to try to force a Photoshop mindset on LR.
    Andrew Rodney
    Author "Color Management for Photographers"
    http://www.digitaldog.net/
  • canoflancanoflan Registered Users Posts: 168 Major grins
    edited June 21, 2007
    Thanks so much...
    arodney wrote:
    There are workflow and quality advantages to doing all the heavy lifting in LR using the linear encoded, high bit, wide gamut pipeline from Raw data.

    Photoshop is great at correcting rendered images where you basically need to change the existing baked in numbers that are represented by (often) million of individual pixels. This isn't how LR operates with raw data (using existing rendered images, its questionable if you should do anything other than use LR as a database although the jury is out). For raw, its not anything like how images are manipulated as we've done in Photoshop for 17 years.

    Both Thomas Knoll who wrote ACR and Mark Hamburg, the chief architect of LR have attempted to place the controls to best handle this raw data and its just different than working with pixels in Photoshop. Its important not to try to force a Photoshop mindset on LR.

    I understand exactly what you are saying here and this helps a great deal. Thanks again for taking the time to explain.
    Pat
Sign In or Register to comment.