I thought I'd change gears here for a day and talk seriously about post processing in this day and age.
We did a photo assignment last Thursday which generated nearly 3,000 large (24 mpxl) raw files. The project was to photograph six individual models in white lab coats. The brief called for the models to be "animated/excited." The "take" needed to be edited down and then quickly shared with the client. The client made 36 selections yesterday and I spent the better part of an eight hour day enhancing the files and dropping out the light gray background we used.
The photography part of the equation; the actually pushing of shutters and positioning lights, is probably old hat for most of you. It was a straightforward exercise in lighting with soft boxes and electronic flashes. The one addition, driven by the need to preserve detail in the lab coats, was the constant use of a "net" to block some of the light falling onto each model's shoulder. The shoulder closest to the main light. Nets are like neutral density filters for the light coming to the model. They can be feathered and they are constructed so that there's no shadow line on the subject. Useful for controlling how much light you want on specific parts of a subject.
The stuff that interested me this time was the post production. As soon as the shoot was over and I was back in the studio I did a very quick edit in Lightroom and dumped about 1,000 files that were flawed in one way or another. I converted the remaining 2,000 files from raw to Jpeg so I could upload the Jpegs to Smugmug.com. I didn't need outrageous file sizes or ultimate resolution so I made the Jpegs 3200 pixels on the long side with a Jpeg quality setting of 90. The iMac Pro I bought 15 months ago whipped the files out fairly quickly and the upload was snappy and uneventful.
I like presenting galleries on Smugmug to clients for a couple of reasons. If I save the files at their full size and highest quality settings (still as Jpegs...) I can use the resulting galleries as a third tier of my back-up. The client is able to access their gallery from anywhere. Multiple people on the client side can use the gallery at the same time. This would allow a large company, with locations in multiple cities or countries, to get buy-in from associates in far-flung offices. I can set a maximum download size and let clients download files individually. They like this because they can use the images for placement in comps while figuring out exactly what image and which cropping would work best in a layout. I can assign a separate password for downloading privileges so marketing people can decide who has access to shareable content.
The gallery of images for Thursday's client was up and running on Friday morning and they quickly selected the 36 images included in my bid for post processing. That's where the real story starts for me about post production...
Here's the way I did it. If you have a better workflow, please share it!
I open the raw files one at a time in PhotoShop via the Camera Raw interface. In that interface I make as many corrections and adjustments as I can. These include fine tuning color temperature, opening up shadows, creating a custom curve to hold onto highlights in white lab coats, and cropping the image into a usable size.
When I finish doing those things I used the new "Enchanced Enlargement" tool just introduced in the latest update to PS. This doubles the pixel count while using A.I. to refine and sharpen details. What I end up working on with this additional processing are 140-200 megabyte files in 16 bit.
Once the files are open I make any additional color, tone or retouching corrections and then I go to the selection tool and click on it which opens two buttons in the top menu bar. These are "select subject" and "Select and Mask." I hit the first one, "select subject" and the program does a very good job (if I've shot everything correctly...) selecting the actual subject and rejecting the background. I can see by examining the "marching ants" outline that the tool either worked or failed. Once I've examined my outline selection and am happy with what it has selected I hit the "select and mask" button this uses the subject selection to create a good, tight mask which separates the subject form the background.
This part of the process is rarely perfect. There's always some things that gets included but should not or something that gets masked out (the tops of ears! the ends of fingers! ) that very much needs to be included. This is the part of the process where you get to go in and try two different fixes to get your file selection exactly as you want it. When I'm in this stage of the process I like to work with the image at 100% and carefully check every edge. 100% on a file that's 12,000 pixels on the long end is quite a close range observation but you catch more artifacts and glitches this way.
The first of the two tools is the "refine edge" brush tool. You can change the diameter and use the tool to smooth edges and make small adjustments to what stays in and what goes out. I like to use this for hair as you can make sure you've selected the hair from the background so it looks realistic when the client drops the image of a person onto a new background. I use it around edges that might look a bit rough due to the initial selection process. It's a great tool and the associated menu gives you an almost infinite amount of control for edge detection, feathering, contrast and other fine-tuning parameters. Light touch or heavy touch? It's totally up to you. The tool works best when there is a good contrast and color difference between the background and the subject. The more alike the foreground and background the worse the performance.
The second tool is a brush tool which adds to or subtracts from a selection mask. I might have part of a white sleeve that doesn't get selected along with the subject. Since the mask visualization is 50% opaque ( you change the percentage) I can see detail under the mask that may need to be included but is currently outside the mask. I use the brush tool to wipe over the desired but excluded detail and it erases the part of the mask blocking the detail from showing up. If I've included details I don't want on the edges I can hold down the option key while brushing and it adds to the mask.
Once I think the selection and masking are optimal I can choose a number of ways to save the image. Since I'll be delivering these to a graphic designer who will be compositing them into a background she'll be creating she'll want the image with the background dropped out and the subject floating in a transparent layer. I select "new layer" and PS adds a layer with the subject visible and the background dropped out above the original background image which is how the image looked after the raw ingestion but before the selection process.
I save the completed file in layers as a 16 bit Tiff. The file sizes of the finished 36 files ranged from 300 to 400 megabytes apiece but they are big, sharp and easy to use. The client will most probably convert from 16 bit to 8 bit but might want to stay in 16 bit if they have some additional, tricky color work they'd like to do on the files.
The time consuming part of each image refinement lies in the fine-tuning of the selection edges before committing to the final step of outputting the file to a layered format. I can spend half an hour getting the hair and edges of an image looking just right but I have to say that it beats the heck out of the "old school" way of selection in PhotoShop, using the pen tool to make point-by-point clipping paths...
On projects like this one which include bigger rights packages (yes, that's still a thing...) I like to give my clients a folder of the finished Tiffs as well as the original raw files (in case I get hit by a truck and they still need to use the files) and the intermediate .DNG files which are created in the Enchanced Enlargement process in Camera Raw.
While I have to give up a bit of control, giving the client the raw material means fewer panicky phone calls from folks looking for different files or wanting to create different looks under super tight deadlines. They are generally as good at post production as I am and can grab the file they want to do their own post processing. Post processing isn't really a profit center for me; I have fun and make my money shooting the images and doing the production. The post production stuff can be creative but I don't have the mindset to make it too complex and creative. I feel sometimes that it's like washing the dishes and pots and pans after cooking dinner. It's not as much fun as cooking, far less fun that eating a good dinner, but a necessary part of the whole project.
A few things about shooting with the intent to drop out backgrounds:
1. Working with higher resolution files works better than working with lower ones.
2. Automated selection tools work better with files that are in sharp focus.
3. Soft focus or soft edges can be problematic with automated selection tools.
4. My highest success rate comes from files shot at f8 to f16.
5. Flash is a good idea for this kind of work because it freezes most motion.
6. Automated selection tools are almost never a good match for images with motion blur.
7. The "refine edge" tool hates soft edges. The brush tool for adding and subtracting masked
areas is ambivalent about soft edges.
8. Nailing focus is important and the corollary for me is the discovery that eye detect AF
doesn't work as well with subjects who are jumping up and down. Best to make D-o-F your
friend on these kinds of shoots.
9. Take some extra time to fix stuff on your set, like burned out white areas on a medical
white coat. It will make life easier in post production.
And, finally, be sure you have the life affirming elixir we call "coffee" at hand when you need to take a break, walk away and find, once again, a reason to go on....
So, after swim practice yesterday I spent most of the day huddled down in my office; mouse in one hand, keyboard under the other, grinding away on the images I'll be handing off tomorrow. And that's enough work for this month...
