Recently, I run across an interesting article on how to convert color digital photos to B&W. It seems a direct approach yields cerulean skin and other undesirable features that are absent in traditional B&W portrait photography. The process is a bit involved, and I am writing this post so I don’t forget the details. I cannot say I understand all the steps, either!
First, legal matters. The information is taken from Photo Práctica Actual, a Spanish magazine, issue ????, year 2008. They cite original ideas from ???? (see ???).
I experimented with some photos of my daughter, with impressive results, since she is blond and the background was not as dark as one would have wanted. But, I am reluctant to publish this stuff, and I looked for material with the right copyright. I am using a cropped version of the “Gothic” portrait I found in wikipedia, by Stephen Paul West. Results are quite good, perhaps not so nice because the original is a bit dark already (Gothic, in fact). These are the steps to be taken in PhotoShop (I am pretty sure the same can be done with other programs, like the gimp):
- A quick conversion to B&W via PhS or the gimp (“desaturate”) would produce the image in the middle, ok but far from B&W portrait greatness.
- Set depth to 16 bits, then “Lab” mode. Select the “L” chanel, then set this one to B&W, neglecting the other two. This is a well-known general procedure to obtain B&W images (not just portraits). The “L” stands for lightness , see the wikipedia entry, and it’s supposed to be the “best” channel to keep when going from color to B&W (another firm candidate is often the red channel, the “R” in “RGB”.)
- Select the “darker” areas. In PhS, the simplest way is to ctrl+click on the tiny thumbnail in the channels menu, which will select the “brighter” areas, the invert the selection. (I still don’t know how dark these regions are, and how to change them.) Then, insert an new fill layer, filled with a solid black color (0,0,0). This dramatically increases contrast in the dark regions, and darkens them. The layer mode can be kept as “Normal”, but the transparency should be lowered to about 50% or so for this effect not to be overwhelming.
- Perform some contrast enhancement, by e.g. inserting a new effect layer with “curves.” (this is also implicit in step 1).
- This is perhaps the most technical step. To add some grain effect to the highlights, we need an extra layer. Duplicate the background layer, then use a “high pass” filter on it ( filters -> other -> high pass). The standard 10 pixel radius seems to be fine. The resulting layer should be added as “strong light”, or similar, and its transparency adjusted to about 30% or so. To produce a jpg, remember to go back to RGB mode (merging all layers), and 8 bits… manually (I think this is automatic with the gimp).
Above I have inserted the original, straight (step 1), and final (step 5) versions. I include the intermediate ones below.