cgardner
Offline
• • • • •
Upload & Sell: Off
|
The apply image trick is one I learned in the Dan Margulis "Photoshop for Professionals" book. He has a second on Lab processing which I have not read but understand is quite good.
RGB filters are used on a digital sensor because equal parts of RGB light create neutral gray tones, but RGB isn't very good at recording detail in objects with strong RGB or CYMK colors due to the way the color is split into the discrete channels by the filters over the sensor sites and how that process differs from the way the eye senses color.
RGB doesn't mimic the way the eyes perceive color. The cones in the eyes sense blue/yellow and green/magenta, not RGB. The rods in the eyes, which cover most of the retina, are sensitive to narrow band of greenish light and are 3000x more sensitive than the cones. The sensitivity difference limited spectral sensitivity of the rods is the reason why a Bayer RGB sensor needs twice as many green sensor sites as red and blue.
Photoshop and Lightroom use CIE*Lab as the "translation" space. Created back in the 1930s the Lab model of color is based on the gamut humans can see with their eyes based on how the cones and rods of the eye work. Luminance carries the detail similar to how the monochromatic rods record what hits the retina: as a B&W image. The a and b channels record the relative amounts of blue/yellow and magenta / green similar to how the cones of the eye actually sense color.
RAW files and JPGs are also encoded similarly as Luminance, blue/yellow, green/magenta components then decoded back into RGB to drive the RGB pixels on the monitor. When you open a RAW file in ACR or Lightroom and pick an editing space (sRGB, AdobeRGB, ProPhotoRGB) the discrete colors defined by the number of bits at capture are assigned both an absolute set of Lab coordinates based on the size of the editing Gamut.
To visualize how this mapping works create a file with solid blocks of RGB and CYMK then convert the file from sRGB, Abobe, ProPhoto.
The RGB values always stay the same and the blocks will look same on the screen, but if you open the color picker you'll see that the Lab values change.
In Photoshop you can switch image mode back and forth from RGB to Lab without affecting gamut because the Lab is the translation space. That's what allows Apply Image to overcome the limits of RGB capture in primary and secondary colors.
Yellow is R+G in equal amounts. In a bright yellow flower most of it will be almost fully saturated 240+ values of R and G, with darker shading recorded in the only remaining channel, B. The B "grays down" the R+G yellow. Magenta and Cyan objects are similar in that two channels carry the color leaving only one remaining to provide most of the clues about 3D shape and texture. Reds, blues an greens similar to the colors of the filters on the sensor can also lack detail but in those colors there are two other channels (e.g. G and B for Red) creating the shading.
Because there isn't much blue light reflecting off a yellow flower onto the sensor the camera does a pretty lousy job of recording the same amount of detail your eyes see in the flower. As a result a photo of Daffodils or Sunflowers always looks flat for a reason that isn't easily grasped — unless you know all that color theory stuff above, which I learned managing offset printing for a living 
In an Lab copy of a Daffodil the detail is carried in the L channel as a B&W image. That detail can be enhanced in the flower by finding the channel in an RGB copy of the flower which has the most shading details (Blue) and adding it to the L channel of the Lab copy...
This is an unusual application of the technique but the only one I have on line which shows the Apply Image user interface:
The person I was helping in that case wanted to make the girl's freckles more prominent. Looking at the RGB channels the freckles were most prominent in the Blue channel because skin is predominantly R+G with B creating the darker shading. What the screenshot shows is how the detail in the Blue channel was copied into an Lab copy of the same file. Here are the steps if you want to try it:
1) Start by opening the file and saving two separate copies in .PSD format: Keep one in RGB, but before saving the other change the mode to Lab.
2) Open both copies. Make the Lab copy active and open the channel window.
3) Click on the L channel - that select it as the destination for the applied data.
4) Hold down the command key and press the tilde key ~ That will change the view on the screen from just the L channel to the entire color image, which allows you to see the effect of applied layer.
5) As shown in the screen shot select the channel with the detail in the RGB copy as the source. The target layer L was set in step 3. The mode menu controls how the Blue channel info is added to the L channel. Overlay adds the two numerical tonal values together. As shown in screen shot you need to adjust the % of the Blue that is applied by eye until you get a result you like. Also try different modes such as multiply.
6) Once you get the color to taste in the Lab file you convert it back to RGB.
In the case of a Daffodil doing the Apply image to add more detail to the yellow flower will also affect all the other content of the photo. So the next step is to mask out everything except the flower in the edited version. There are lots of ways to do that in Photoshop. I usually just select all in the edited Lab>RGB file, copy, then paste as a new layer in the other unchanged RGB copy creating a new layer on top. I then use the magic wand or selective color to isolate the flower and create the mask on the top layer.
Once the edited version of the flower is isolated with the mask its just a matter of adjusting the % opacity on the top layer to blend in the "detail on steroids" into the original. To nuance it even more selectively shade the mask on the edit layer.
Here is a shots previously posted on the forum I edited for the OP using the technique but shared via PM:
This one was also posted previously...
One of the advantages of CS5 vs LR is the abilty to create layers and mask to apply corrections selectively. The masking tools in CS5 are also much improved over earlier versions of Photoshop.
The value of learning as many PhotoChops as possible is that it allows you to see past what your eyes are seeing to understand before you take a photo what the final image, with manipulation, can look like.
Working in printing I've used Photoshop since version 1 in the early 90's years before ever owning a digital camera. Its been part of my "creative DNA" (to the extent the son of an engineer and has any) as long as I've been shooting digital. Before digital, back in the days when I used the zone system and B&W film to manipulate reproduction, I also used yellow or red filters to darken sky, or a green filter to all more contrast and "pop" to foliage. I also don't recall ever making a print where I didn't also do some burning and dodging to alter the mid tones and guide the viewer to the focal point.
|
