noise to signal:
Digital imaging

Photosynth

It's always exciting when I can finally talk about, and show people examples of, something that our research team has been working on for a while. In this case, the very cool Photosynth.

From the Photosynth site: "Photosynth takes a large collection of photos of a place or object, analyzes them for similarities, and displays them in a reconstructed 3-Dimensional space."

One of the interesting things here is that images of any resolution or angle are useful as part of the data. Crappy camera phone image of a detail? That might be more detail than is contained in any of the other photos, so it might be useful for zooming in. Accidental shot of the ceiling when you bumped your shutter relase? Also a valuable part of the data set.

Still at it

Here's one of Six Flags for Joe.

Update: Vince suggested that a less blurry version might be better, and he may be right.

Fake real model photography

Remember the super cool aerial photography I linked to a couple of weeks ago?

One of my co-workers found a great article talking about how to do the same thing in post: Fake Model Photography.

And since we're all enjoying the Bird's Eye imagery in Windows Live Local, it seemed a natural step to use that imagery to experiment on.

My first attempt, with this set of little boxes, worked pretty well:

But my second attempt, using this tasteful little place, worked a lot better:

Greg is doing some fun ones, too: Lady Liberty and a super awesome Doughnut.

Update: Here's a nice image I did of Columbia University.

How about some neat smokestacks? (NYC folks -- any idea what they are?)

The WhiBal

One of the great things about digital photography is how many choices it's given us that we didn't have before. A good example is film speed: With film, unless you're shooting with certain specific kinds of cameras (or you're willing to suffer tiresome workarounds), once you put a roll of 100 speed film in, you're basically stuck at that speed until you've finished the roll. With digital, if you go from a bright situation to a dark one and back again, you can switch film speeds to your heart's content.

You can do the same thing with color balance, too. Briefly, for those of you who don't know about color balancing: light comes in different colors. Straight sunlight is white; sunsets obviously are more orange. Cloudy days are much more blue, indoor (tungsten) lights are very orange (which you've noticed if you've taken pictures on film indoors without flash), and flash is slightly more blue than daylight. Professional photographers doing "color critical" work on film will use an elaborate system of color meters and filters to balance the light before it gets to the film, and work hard in the darkroom to make sure the color is exactly right (or, in my case, as right as I was willing to spend the time for).

In digital, you can do the same thing with color balance you can do with film speed. Outside? Set it to daylight. Inside? Set it to... uh... whatever inside is. A little lightbulb, usually. I can't remember. I don't really pay attention much because I always shoot in RAW mode, rather than JPEG, which means that I can easily change the white balance after the fact. I open the RAW file in Photoshop, choose the color balance I want, and... it's still not quite right.

This is a fundamental problem with color balance: the presets are never quite right. Tungsten lights change color over their lifetime. Fluorescent lights are all over the map. Sunlight with a bit of cloudyness is going to be slightly blue. And so forth. Now, in Photoshop, you can tweak the color balance to your heart's content, but that takes a long time.

The solution, and the point of all of that long, overwordy introduction, is to have a neutral reference point to balance off of. And the excellent tool I've been using for that for a few months is called a WhiBal.

You can follow that link to learn everything the manufacturer wants you to know about it, watch the video demos, and find ordering information -- it's about US$40. I'm just going to show you some examples of how much of a help it is. And no, I'm not getting a kickback from them, although I've sold enough of these for them I should (hint, hint, Michael :-)

Here we have a scene lit by sodium vapor lamps, which may look vaguely familiar. On the left is the original scene; on the right is my reference photo of the WhiBal card.

And here are the corrected versions, sampling off of the WhiBal card.

Now, I could have corrected it by hand to get that color -- but this is so much faster and easier that there's really no point.

Here's another example. These daffodils were on our kitchen counter, under our tungsten lights that have yellow glass shades. On the left, obviously, the uncorrected one. And remember, the entire process was "open both files in Adobe Camera RAW, sample off of the WhiBal card, click done." It just about took more time to type that than to do it.

One point I want to make clear is that this isn't just for extreme situations like the tungsten and sodium vapor lights in the previous examples. Here's a picture of Rusty I took outside our house. On the left, the uncorrected one looks OK, really -- a little blue, maybe, but still basically fine. The corrected image on the right, though, is a much richer and more accurate image.

So, go check the WhiBal out. It's a huge timesaver and a great tool.

A peek behind the curtain

The title could describe the photo I'm posting today, which I realized only after I wrote it, but what I really meant is that I thought this would be a good picture to use as a another demonstration of what a photo goes through here at Noise to Signal before it gets posted.

I have what could be described as a manipulative printing style — what we might have once called "heavy on darkroom technique". (And yes, that sentence was designed to make my friends still working in the darkroom cringe, because I'm a bad person.)

Nearly all black and white printers manipulate contrast and burn and dodge to some extent, but more manipulative printers treat the original image as no more than a basis to start from. This doesn't mean that we use it to cover for poor technique at exposure time — in fact it's much easier to be a manipulative printer if you're starting with a pefectly exposed image. As Ansel Adams (himself an extremely manipulative printer) used to say, "the negative is the score, the print is the performance".

My own printing style is heavily influenced by a photographer named Roy DeCarava, who is a master of getting beautiful, rich, expressive detail in the shadow areas of his prints, with very few light values in the photo.

So, on to the demo. Let's start with the raw image, straight out of the camera with a default conversion applied. You can see that it's pretty bland, with none of the richness I'm looking for in a final image. Plus, hey, it's in color, which isn't what I want for this photo. (There's a whole discussion here about how I still see mostly in black and white that will have to wait for a future post) There are, however, the bones of an interesting photo here.

A default conversion to black and white looks a little better — it's starting to get more graphical, which is what I'm looking for — but it's still pretty gray overall.

Increasing the contrast while converting it to black and white with The Imaging Factory's Convert To BW Pro brings it closer to what I envisioned when I saw the photo originally, but it's still not quite right.

And here's the final version. I've increased the contrast on the flag in the window, increased the contrast on the bricks on the left side of the image, and added some edge burning to the left, right, and bottom edges. In the web image the detail in the shadows to the right of the flag seems to have disappeared, but there's plenty hinted at there in the print version.

And, because I can't resist, here's another "that 5d captures so damn much data I can't stand it" crop. What's that in the window below the one with the flag in it? Why, it's a dancing Shiva (?) and an elephant tucked in the corner. I could probably extract a bit more sharpness out of these, but I don't really want them to call too much attention to themselves in the final print.

Possibly too much data

Canon's begun shipping a new camera, the 5d, which I'm seriously thinking of upgrading to. I've been reading reviews and looking at samples, and recently I found some samples from the 5d on Canon's site.

They're the usual weirdo sample pictures that camera manufacturers post. Generally they're designed to show off some quality of the camera -- for example, the nice rendition of the subtleties of the whites in the dress, below.

I noticed something when I opened them up in Photoshop, though. Once you're talking about 13 megapixels, with a great lens at it's sharpest aperture, you may see detail that the grain structure in film, or a lesser lens, might have obscured before.

For example, this model in a bridal gown looks lovely, and the whites are rendered nicely indeed...

... but she also needs a little more cover on her chin.

Another model shows off how nice the out-of-focus areas look ...

... but the in-focus areas show that the tip of her nose is dry, and her pores could be a bit cleaner.

The Imaging Factory's Convert to BW Pro 3.0

I'm not much in the habit of recommending photo tools, but for those of you doing black and white work in digital more than occasionally, I have to recommend The Imaging Factory's Convert to BW Pro 3.0.

Now, normally I don't go in for most of the filters and whatnot that are available to do stuff to digital images -- either I find them effects tacky, or I'm good enough of a Photoshop jockey to find the redundant, but this is an exception. I had been happy with the techniques I was using to convert color images to black and white -- pure channel mixing by hand, or Rob Carr's excellent method he developed for Greg Gorman. And I'm not sure that I couldn't get the results I'm seeing in Convert to BW Pro using those techniques -- but this is a lot easier, and works a lot more like the way I think than those other techniques do.

By way of an example, here's an image you've seen before, in the original and converted using their tool. I did no burning or dodging; the only thing I did outside their tool is to apply a bit of a split-tone effect. Like I said, it's not impossible that I could have done this without their tool... but it was much faster and easier. It's definitely worth the $99 to me, and definitely worth the 30 day trial to you to try out.

Check them out: The Imaging Factory.

(And thanks to Michael Reichmann for pointing me to it in the first place)

Split-toning in Photoshop

This is another one of those techniques that I was sure everyone knew about, but given that I've gotten a couple of questions about this picture of Rusty, I figured I should talk about how I did it.

The term split-toning comes from the wet darkroom. I'll quote the excellent description in this article on split-toning from Handcolor.com:

Split toning is when the toner acts only on certain areas of the print, the middle or low values, while leaving the rest of the print with no color change. The old Agfa Portriga (particularly in the matte finish) would often turn a beautiful purple-brown in the low values, while the rest of the image would remain unchanged. This resulted in prints with much greater apparent "depth."

I'm starting to really like the digital equivalent when I post black and white photos, for much the same reason. Pure black and white images are often a little washed out onscreen, especially when they're shown in context with color images. I suspect this has something to do with the way images are displayed on computer screens, but I don't really care enough to investigate. Also, prints (whether silver gelatin or inkjet) are never truly black and white -- they all have a little tint based on the emulsion, paper, and/or inks.

This technique would also work well for those of you printing black and white images on color printers.

OK, let's get to it. First, here's our regular black and white image. I've deliberately chosen one with strong contrast, to make the samples more obvious, but you can get beautiful results with much more subtle images.

The way I like to add color to a black and white image is to create a new color layer, fill it with the color of my choice, set the blending mode to multiply, and drop the opacity of that layer down. Here I've used #FF9000 and an opacity of 16%. I like the way it looks in the shadows, but the highlights look a little tacky to my eyes -- like they're behind a sheet of yellow glass or something.

Dropping the opacity of the layer to 6% makes the highlights look better, but now I'm not getting the richness in the shadows I liked before.

This calls for a special technique. First, let's bump the opacity of this layer back up to 16%. Now, on the layers palette menu, choose "blending options". In Windows, you can also find this on the right-click menu if you right-click on the name of the layer you want to affect.

Down in the bottom-right corner of the dialog box that comes up, you'll see a pair of sliders like these:

We're interested in the bottom one, here. What this slider does is change the transparency of the current layer based on the layer below it. Any gray tone that appears between the sliders is opaque; any gray tone that appears outside of the sliders is transparent. It's easier to understand if you fuss with it yourself, but here's an example: I've dragged the highlight slider to the left, so anywhere the layer below the current one has a gray value brighter than 140, the current layer is transparent.

As you can see, this gives kind of a patchy effect, but I'm getting closer to what I was looking for: the shadows are getting the tones, but the highlights are being left alone.

Now let's get tricky. Take a close look at the slider we just moved. Notice how it's got a line down the middle? Almost like you could, oh, I don't know, maybe... split it in two? Bingo.

Hold down the alt or option key, click on the right side of the white slider, and drag it to the right. Zout alours! The slider has split in two.

What happens when you split the slider is that it smooths out the transition between opaque and transparent in the current layer. Where before the pixels were either on or off, we now have a nice gradient: pixels below 106 are opaque; pixels above 170 are completely transparent; and between those two values the pixels are some intermediate step. This gives us a much nicer looking image:

Instead of the abrupt jumps between toned and not toned, we now have nice smooth transitions. My highlights are largely untouched, and my shadows have the richness I was looking for. Success!

You can, of course, use this same technique to end up with an image that's much colder in the shadows, too:

Here's the original again so you don't have to keep scrolling back and forth to compare:

I hope that turns out to be useful for some of you.

If you want to get really tricky, you can tone the shadows one color, and tone the highlights an entirely different color... but I'm going to leave that one as an exercise for the student.

Photoshop cross-processing

It constantly amazes me that while much of the content on the web that providers want people to pay for is mediocre at best, some wonderful things are completely free.

Case in point: Shan Canfield's wonderful cross-processing actions for Photoshop, which I just discovered yesterday.

Cross-processing, briefly, is a photographic technique where (usually) slide film (E6) is processed in the set of chemicals usually used to process print film (C41). (Less common is cross-processing in the other direction -- print film processed in slide film chemicals) The result is images with oddly skewed colors and increased contrast and saturation.

This technique was a bit overused in the late 1990's, but it can be quite effective when used in the right situation. I like using it (gently) for my photos of Circus Contraption -- the slightly odd look to the colors works well with their style. It can also be used for motion picture film; for example, in the movie Three Kings, most of the desert scenes are cross-processed.

The problem we're running into now, however, is that this process depends on shooting with film. What happens if you're one of the photographers who have made the jump to digital?

That's where Ms. Canfield (also known as Photoshop Mama, according to her site) comes in. She's produced a set of very nice Photoshop "actions", or macros, that do a very credible job of mimicking real cross-processing in a digital tool.

Unfortunately Ms. Canfield's site uses frames, so I can't point you directly to the cross-processing section, but if you go to her site (opens in a new window), click on "shanzcan tutorials", and scroll down to number 23, you'll find them.

"That's all very well and good, Dave, but can you show me some samples?" I'm glad you asked. Here are some samples using the old Photodisc color calibration test image. First, the unmanipulated image:

There are three versions of the action. The first one focuses more on the hue and saturation changes cross-processing creates. The second focuses more on the contrast changes. Here they are:

The third action combines the two:

The really elegant thing about how Ms. Canfield has set up these actions is that the the actions generate cross-process layers that sit above the original, unmanipulated image, and affect it. In the case of the combined action, it creates two layers, one for each of the other actions it includes.

This suggests a few ideas. First, I've found that it's most efficient to just run the third action. Like real cross-processing, the effect on an image is sometimes hard to predict. Having both versions there allows you to turn one or the other off and choose which one you like better -- or change the opacity of one or the other, if you want some of the contrast effect, but not quite as much as her action gives you, for example.

The second idea this suggests is that it's possible to duplicate one or another of the layers if you want more of that layer's effect.

The third idea, which follows on the second, is that it's possible to add other layers to manipulate the effect.

On the left, below, I have a version of the image where I have two layers of effect #1, and none of effect #2, for a stronger hue shift.

On the right I have a version with only one layer of effect #1, plus a color balance adjustment layer (0/50/0 in the shadows and 0/0/25 in the highlights) that mimics the way cross-processing can sometimes make the image go green in the shadows, and warm in the highlights.

One caveat is that because these require layers, they won't work on 16-bit images just yet. But Photoshop CS is right around the corner, which finally allows most features to work in 16 bits. Huzzah!

The possibilities are vast. I'm sure this is giving all of you Photoshop types some great ideas. Enjoy! And mad props to Ms. Canfield, as well.

One for the ladies

Sisters, ever look at photos models in magazines and despair at ever looking like them? Ever think that they have impossibly beautiful skin? It turns out you're right: it is impossible, in the real world. But not with Photoshop...

Greg Apodaca spills the beans, showing a couple of re-touching examples that ddemonstrate just how far from "reality" many of the magazine shots you see really are.

Take a look at example 1 and example 2.

Pushing pixels

There's an article in the latest issue of Lenswork by Frank Van Riper, where he talks about building a great portfolio print by print over one's lifetime. It's a good, inspiring article.

I like to read Mr. Van Riper's essays, in Lenswork and in the Washington Post online -- he wrote one about the Leica M series that perfectly captures my love for that camera -- but I have to address something he mentions in the current article in Lenswork.

He says:

"... here I could not help but think of being in the darkroom, producing archival black-and-white prints on actual paper from actual negatives -- immutable, permanent things, not computer-generated facsimiles of questionable provenance born of a digital camera's bits and bytes.

... it's the very "virtualness" of digital photography that bores me and gives me a headache. The ethereal, never-quite there quality of it all..."

Now, first of all, breaking from his otherwise great essay to engage in some unprovoked digital hatin'... that seems unnecessary. In other essays, he talks about using a Canon Powershot G1, and even describes it as "this terrific camera", so clearly he doesn't reject digital completely.

What's more interesting to me, though, is the content of his comment. I don't want to deny his experience -- if digital isn't real to him, it isn't, and I wouldn't want to try to convince him any more than I would want someone else to try to convince me that Steely Dan makes really good music.

My experience, however, is somewhat different. Digital, to me, has always been as real as anything else. It's not ethereal at all -- it's right there, as physical as I need it to be, coherent and understandable and very nearly concrete.

I suspect the difference is that I grew up using computers, and I have a level of comfort and trust with them that Mr. Van Riper doesn't. I don't mean to imply that I think he's old and out of touch, either -- that's not the message here. The point is that I have a lot of experience using computers as an artistic tool, and that affects my perception of them.

I've encountered this with other designers as well, although not so much lately. When I was first working out in the world, other designers who hadn't really used digital tools would complain that they couldn't place things on a page precisely -- that they felt like they were doing layout with a pointer on the end of a long pole. I, on the other hand, always felt like I had more control working digitally than I ever did with manual paste-up. Pens leak and jump around. Rulers need to be held straight. Wax is sticky stuff, and when it touches a board it tends to want to stay there. Digital gave me much more freedom to be as precise as I wanted to be.

I've seen the same phenomenon, to look at this another way, when I try to explain to non-photographer friends the process of printing traditionally. People whose experience of photgraphy is limited to taking the little metal cannister out of their camera, dropping it off at a store, and getting their photos plus some mysterious orange strips back, do not understand that first you develop the film, then you put it in an enlarger, and only then do you end up with a print on photo paper. It's all very abstract and mysterious. "Why do you need to make it bigger?" "What happens to the film?" etc.

The point of this rambling, I think, is that just because something seems mysterious and abstract doesn't mean that it isn't real: it just means that one doesn't have experience with it.

As Mr. Clarke said, "Any sufficiently advanced technology is indistinguishable from magic." And we distrust and fear magic until we learn to control it.