Chapter 11 - Retouching in LAB - Late and Incomplete

I loved the next to last paragraph in Chapter 11......

The author states, "I have tried to point out some of the areas in which retouching in LAB is superior to more customary alternatives. IF you followed every example, pat yourself on the back, because most experts would not be able to.
For sure, you should not be upset if you didn't foresee blending an inverted B into the L to kill a moire"

With that bit of understatement, let us begin our overview of Chapter 11 "The Best Retouching Space." I have broken it down into our own series of chapters.

CHAPTER 1:

I volunteered for Chapter 11, because I wanted to see if I could make a sundog visible in an image.
When I shot the sundog on the way to Mackinaw Michigan last fall, I could only see the ends of the sundog, but was unable to see the any trace of the arc in the sky, with my naked eye. The arc was totally invisible.

Nevertheless, I KNEW it was there because I could see the ends of the arc, and I had read Galen Rowell's description of sundogs, which are arcs of colored rainbows about 22 degrees either side of the sun. I shot several exposures to see if I could capture it. The sundog was not visible on any of the original RAW images. I wanted to see if I could edit it in Photoshop, to make the invisible sundog, visible.

That is what seduced me to agree to review chapter 11: "The Best Retouching Space"

It is a more difficult chapter to thoroughly understand, than I envisioned.


Chapter 11 opens with a general discussion of what is retouching and the author says it is one of several things.

1) Erasing dust, hairs, scratches and other undesirable elements, such as an ex-spouses faces

2) Putting things into an image that were not there previously eg: inserting a rhinoceros into an image of a cocktail party, or filling a blank area with new bogus detail

3) Completely altering the emphasis on an image element, as blurring or fading the background, or turning part of an image to Black and White.

4) Color correction whether global or local.

For the first task, de-dusting et al, the author suggests any color space is equally as good or as bad.
In the mid-90s, retouchers used a Quantel box that worked in HSB color space, which is a considered a variant of the LAB color space, in preference to CMYK pre-press systems in use at the time.

Today, for removal of bigger items, the author suggests CMYK, as the black channel isolates detail where it can be removed more easily, than in LAB or RGB. For introducing new items, LAB may be a better choice, because the LAB space tends to produce smoother blended edges of selections that are altered, than selections in CMYK or RGB, particularly if dealing with altering brightly saturated colors.

Which brings me back to rainbows and sundogs.

If you are going to select a rainbow for improvement or augmentation, where do your draw the edges for your selection??? The author quotes Herman Melville - "Who in the rainbow can draw the line where the violet tint ends and the orange tint begins?" His answer, of course is NOBODY - (except those little demons in LAB who can seem to do it in Photoshop).


All rainbow pictures are candidates for retouching because cameras do not capture the colors of the rainbow as impressively as we humans see them in real life. This has more to do with human perception than digital sensors, of course, as has been discussed in this book all along. But how do we select only the rainbow, and not the foreground or background?

Here is an image of a rainbow I grabbed along the side of the road on the way to Yosemite late in the afternoon.

First, we create a duplicate layer with a ctrl-J as per usual editing in PS. Then, on the upper layer, we steepen the A and B curves almost to vertical and create a strongly over-saturated color scheme.

We then click on the box with a circle in it at the bottom of the layers palette
, hit the "D" key, to bring up a nice solid black for the foreground, and then use a nice, soft brush at a mild 10-50% opacity in the Normal Blend mode to paint away the saturation everywhere but the rainbow. The brush must be in Normal Mode and at ~60-75& opacity. We then blend this layer with the background layer, at an appropriate saturation level to taste with the blending slider on the upper right of the palette box. The Opacity Slider is here is in the Layers Palette well on the upper right.



Unlike Melville, WE end up deciding where the rainbow begins and ends in this adjustment layer step.



Now that we know how to retouch a rainbow, let's try an invisible sundog. This is the image of the sundog straight from Adobe RAW Converter.

The red circles enclose the ends of the sundog - that was all I could see that morning in the sky.


We again duplicate the background layer and radically steepen the A and B curves and then use an adjustment layer to paint away the A and B steepening everywhere but the arc of color of the sundog.
Well, what do you know??!! IT IS VISIBLE after all!


Not terribly attractive, but as an experiment, rather than an artistic exercise, I think this is a quasi-success, making visible, what was previously invisible to the naked eye.

CHAPTER 2:

Next on the list of tasks was replacing items or filling holes. Here, LAB has some significant advantages - that if the A and B channels are not perfectly aligned the damage to the image is barely noticeable, whereas in RGB or CMYK if the channels are not perfectly registered, chaos reigns in the image.

Here is one of my images, a great egret who looks like he is going to get lucky.



In the next image, in RGB, the R was shifted to the right 10 pixels, the B to the left 10 pixels , and the G was shifted down 10 pixels - ugh!!


In the final image the image was converted to LAB, and the A channel shifted to the left 10 pixels, the B to the right 10 pixels, and the L was moved down 10 pixels. Not nearly as noticeable as the RGB version of shifting.

The significance of all of this shifting of channels, is that we can fill holes with the A and B channels without them needing to match too closely, we only have to match the L for a successful replacement, like for example green foliage on the upper left corner of this image of an American coot.




If we were to patch or clone in RGB, the patch needs to come from somewhere nearby and the duplication is frequently obvious due to repeating elements of all three layers. But in LAB, the L can be from nearby but the A and B may be from further away and thus create a more natural variation in the image that is harder to detect.

The above is easy to describe, but deciphering the EXACT steps in Chapter 11 for image 11.6 on page 224, took me a while to decode, and should be skipped unless you are following along with the book in hand. I cannot use the images in the text, they are copywrited unfortunately.

The first step, as always, is to create a second copy by typing ctrl-J to create a duplicate layer. Now, the author says select an area of the greenery sufficient to cover the orange lettering, and copy it and place it in the upper layer over the orange lettering. He does this as a selection, and then CTRL-C the patch patch, but a clone brush would also seem to work to me.
What you do is use crop tool to select an area of greenery to the left of the orange letters and typing Return, thus cropping the smaller area, then ctrl-A and ctrl-C. NOW, you have this selected area of greenery on the clipboard, and if you reactivate the history palette on the
second layer and type ctrl-V and the hold down the "V" key, you can move the greenery selected to cover the orange lettering.

Now use the Blend IF sliders to exclude everything that is positive in the A channel - (Magenta being so much stronger than green in the orange letters)
Now, the glyphs - (the shape of the letters themselves) only remain - the color is matched, but we need to clone in some detail from an L channel to replace the glyphs.

Next, we need to blend the the A and B channels of the upper layer, with the L channel of the lower layer using the color blend mode which causes PS to use the color from the upper layer and the L detail from the lower layer, to finish retouching out the orange letters.

I have used a variation of the above technique on the coot image above. I used the clone brush separately on the A and B channels to create a color match with the surrounding waters, and then used a clone brush on the L channel to simulate the waters surface, and created this image


I used the same technique to remove the bright area above and adjacent to the sandhill crane seen here



Here is the image after editing



I see I also warmed the image also while editing it, but the removal of the elements in the upper part of the image would be hard to do with just the clone tool alone. Using the color from the A and B channels, and the detail from the L, is a nice little trick to know.

Chapter 3:

I spent a substantial amount of time trying to duplicate the technique used on the old Russian photograph but was not able to duplicate it . I am hoping Edgework or another retoucher will drop in here and explain the editing of the old, weathered, Russian photograph as I can't seem to duplicate it.



Chapter 4:

The next topic is really a repeat of one covered previously - that is the use of LAB in a color blend mode to add very light color to a blown area of an image - whether a hot spot on a sweaty face or a specular reflection on a golden dome of a Russian Orthodox church.

Here is a image, shot after dark with flash, of Launchpad, that nicely demonstrates blown highlights on the front of his leather flight jacket.


Repair is straight forward. Convert the image to LAB. Then, using the eyedropper tool we sample a 5x5 pixel area nearby to the blown area, and then use a soft brush set in the COLOR mode at a modest opacity of 15-25%, to paint in some color over the blown pixels. LAB calmly instructs PS to produce an imaginary color and then blend it with the existing highlight to create a new lighter color for the area of highlight. This is very fast and easy to do.



I will remember and use this tip! I had a discussion with Baldy a while back about blown highlights on this fellows forehead, and this tip would have been very useful for fixing it.



CHAPTER 5:

Next we move on to using the sponge tool to modify color and saturation of images.

Using the sponge tool directly in RGB is easy, if the area can be selected, but the author gives us a Russian Orthodox Church roof with a highly geometric pattern of red, green and white tiles, and wants to de-saturate the red only. And maybe move the green a little to the green side as well. Without using selection tools, this will be challenging in RGB.
But in LAB, it is straight forward. By using the sponge tool, and saturating or de-saturating the grey of the A or B channel only, we can shift the saturation of the color and the red/green or blue/yellow balances. The author then uses the Blend IF sliders again to limit the areas effected. I find using the Blend If sliders to select skies in landscapes to frequently be a more subtle selection tool, than using the color wand or Color Select tools.

The author rapidly moves on to using the dodge and burn tools on the A and B channels, again to lighten or darken the grey levels of the A and B channels in the image of the monastery at Sergiev Posad.
I found I was confused as to the difference between using the sponge tool to saturate or de-saturate, and the dodge and burn tools on the same A and B channels.

I thought this was not made clear in the text.

Thus, I embarked on sponging and dodging and burning an image of mine of an old white barn to discern the differences.
One good trick the author makes use of, with the dodge and burning tools on the A and B channels, is that if it is overcooked on the duplicate layer beyond salvaging with the opacity slider, he can go back to the background layer to diminish the effect by using the opposite tool on the appropriate channel in the background layer. Neat idea!!


Here is the barn prior to sponging or dodging or burning.



Here I have used the Sponge tool at 50% to saturate the left 1/3 of the A Channel, left the mid center unedited, and de-saturated the right 1/3 with the Sponge Tool 50%. The left side sponged to increase saturation in the A channel shifts Magenta, the De-saturated right side, shifts toward neutral, less colorful in the Magenta shades.



In the next image, I saturated the B channel with the sponge at 50%, and the colors shifted towards the yellow. I desaturated the right 1/3 of the image with the Sponge set to Desaturate, and the colors move to neutral and less colorful.



Next, I Dodged the A channel on the left, and Burned the A channel on the right, shifting the colors to magenta or green.



I, then carried out the same pattern in the B channel, dodging the left side and burning the right side, shifting colors to yellow or blue.




Dodge A and Dodge B Channels moves both to the warm side magenta and yellow




Likewise, Burn A and Burn B moves the colors to the cool side. green and/or blue



These moves can be done on a duplicate layer, using the opacity slider to control their amounts, and can be done on selections made with any of the typical selection tools. Blend IF sliders can also be used to limit the extent of the image affected also.

Here the barn roof has been burned in the A channel at 27% after selecting with a Color Select command


Here the sky was selected with Color Select, and the burn tool was used to burn the B channel


CHAPTER 6:

Next, we move on to eliminating moire again. Typical LAB images have all detail in the L channel, and should have little or none in the A or B channel. Looking at most A or B channels reveals a grey, low contrast, almost featureless grey image. If there is a definite pattern pattern in the A or B channel it is likely a defect in the image, or a moire'.

We learned in Chapter 5 how to blur the A and B channels to get rid of color noise in an image.

Half tone images - images printed on regular paper in newspapers and magazines - consist of numerous small dots of cyan, yellow, and magenta ink. These are not normally visible to the naked eye, but to a scanner or a digital camera they are seen and recorded. Frequently, the pattern of dots does not mesh with that of the scanner and an interference pattern, called moire', is created.
Here the pattern is not just in the L channel, but the A and B also.

The usual way of dealing with moire' is a early Gaussian blur to diminish the half-tone dots to a blur, and then a USM sharpening to regain the borders lost to the Gaussian blur. I have done this frequently when copying images from a text book with my Nikon CoolPix in the RGB color space.

The author says it is better do retouch them in LAB. The magenta pattern is more important than the cyan, and far more important than the yellow, because it it responsible for more ink on the final page. And second, the author goes on to say that we are not expecting an image as good as a typical digital capture, that a certain amount of graininess will remain.

I know that Edgework uses the Inverted B blending in his work. Maybe he will demonstrate it to us here.