Photoblog 2.0

Dream Stairs, photo by Harold Davis. View this image larger.

“Dream Stairs” takes Stair after Escher to further (and possibly psychotic) places. As someone on Flickr said, “This is messing with my head!”

There’s also an issue of leaving well enough alone.

I think it fair at this point that I show the photographs that the Photoshopped “Dream Stairs”, Stair after Escher, and Endless Stair come from. “Dream Stairs” and Stair after Escher originate in this 10.5mm digital fisheye:

View this image larger.

By the way, Julian tells me that he was really worried I would drop my lens as I changed from fisheye to normal wide angle for the shot that is the origin of Endless Stair:

View this image larger.

Stair after Escher, photo by Harold Davis. View this image larger.

Here’s another Photoshopped version of the spiral staircase in Embarcadero Center. This original was with my Nikon 10.5mm digital fisheye. I varied the iterative technique by flipping the image to get the symmetry.

The image reminds me a little of an impossible M.C. Eshcer staircase.

Endless Stair, photo by Harold Davis. View this image larger.

On Friday Julian had a half day of school. I picked him up early, and we went into San Francisco where I took the photo of a spiral stairway at 2 Embarcadero that is the basis for this post-processed digital image of an endless stairway. The post-processing was similar what I used for the endless doorways in World without End.

The stairway at the Embarcadero Center was narrow and in a dimly lit spot, somewhat exposed to the elements. I used a wide angle to make the small space seem expansive. My camera was on tripod, with a long exposure for maximum depth of field. I wanted everything to be in focus so that there was a good sense of depth.

The lighting was very mixed, with a wide variation of color temperatures, predominantly flourescent blue at about 3500 Kelvin and ambient reflected daylight at about 5600 K, but also some incandescent from a nearby fancy watering hole. I’m not sure what the patrons of this restaurant made of Julian and me diligently photographing this dingy stairway, but some of them were certainly observing the spectacle through a plate glass window. And thanks for the rather elegant mensroom, we made good use of it when the photos were “in the card”.

My first step in post-processing was to open different exposures from the RAW, balancing each to a different color temperature light source, and exagerating the cast of the light.

After I combined the RAW conversions, I used a process that is essentially iterative, or maybe even recursive: I made successively smaller copies of the original image, and then pasted it on a layer of the original. Each new copy was smaller than the original roughly in the same proportion.

Once the layers were complete, I archived a copy, flattened the image, and then did minor retouching.

World without End, photo by Harold Davis. View this image larger.

There’s more to this picture than meets the eye. Put another way, no matter how hard you look at the on-screen version, even in its larger size, you won’t see everything that is in the image. Let me explain the alpha and omega of it.

The photo the image is based on started as a time exposure of the corridor in the Officer’s Quarters at Fort Point under the Golden Gate Bridge in San Francisco. (View the EXIF data here.) This was a normally-sized photo taken with my Nikon D200, so at 300 dpi (dots per inch) it measured 2592 X 3872 pixels (about 8.6 X 12.9 inches). I mention the dimensions of the image because they become an issue later in this story.

When I looked at the photo on my computer, I saw six “regressions”–or doorways on the corridor. At the end of the corridor was a dusty, dark, and unattractive display case.

I decided to replace this unattractive display case with a proportionally smaller version of the duplicated image itself, to make the corridor seem to go on longer (for twelve apparent regressions).

It wasn’t very hard to paste a smaller version of the image in, thus extending the corridor to twelve doorways. I duplicated the image, measured the size in the original image I needed, and resized the duplicate accordingly (to about 1/6 of the original dimensions, as it happens). Next, I dragged the small version of the image over the large version to position it in place, and dropped it to add it to a layer in the original image. I used a layer mask, the Paintbrush Tool, and blending modes to combine the smaller version fairly seamlessly into the large version. In addition, I realized that at the point in the process where I flattened the layers I’d need to do a bit more detailed retouching using cloning and related Photoshop tools.

It’s worth noting that as a matter of workflow I had to flatten the layers in my image a number of times. Each time I did this, I saved an archived version of the image with the layers still intact (except, as you’ll see later, when my layered versions exceeded Photoshop’s maximum file size of 2 Gigabytes). I did this archiving at every step of the way where it was possible, just in case I later decide to rework the image, and I won’t note that I did it each time (but you should do it too when you work with layers).

Looking at what I had, the thought occurred to me, why not repeat the process? So I resized the duplicate image once more, down to about 1/6 of the reduced size, or 1/36 of the original size, and added the new, small version in position as a layer to the original image. Once again, I used a layer mask, the Paintbrush Tool, and blending modes to combine the versions, with the understanding that further retouching would be needed eventually.

Now I was up to eighteen regressions, and the corridor with doorways looked pretty infinite to me. But, in fact, the corridor didn’t go on forever. I had to get up pretty close to it, but at 1600% magnification I could see a highly pixelated version of the ugly old display case.

It turns out that 1600% is Photoshop’s maximum magnification, so I couldn’t zoom in any closer. In addition, the corridor-end-within-a-corridor-end-within-a-corridor had shrunk to an area of about 12 X 18 pixels, so the physical limitations of the building blocks of my image seemed to prevent adding any more regressions. For those of you who don’t know from pixels, this is an area smaller than the proverbial pinhead upon with angels dance. As the Wikipedia notes, “A pixel (short for picture element, using the common abbreviation “pix” for “picture”) is a single point in a graphic image.”

Looking at this 12 X 18 pixel canvas at the end of the known world, I figured why not write something on it. True, no one would be able to see what I had written without being able to examine a full-size version of the image in Photoshop or comparable software at high magnification, or perhaps by vliewing a good, large print up close with a magnifying glass. But it seemed worthwhile to embed a secret in the corridors of apparent infinity.

Typography at this small scale proved to be surprisingly difficult, and I had some much appreciated help from Phyllis. I can’t tell you how much it helps my work having a spouse who is also a graphic design whiz!

My first thought was to write my name, but “Harold” simply wouldn’t fit in the space I had, even with the strokes only one pixel wide. So a single letter would have to do. In a representation of infinity, perhaps α (alpha) or ω (omega) would have been most apt, but I settled for the more egocentric “H”. (I am, after all, at the core of my own solipsistic universe!) As I’ve said, you’d need to view the original Photoshop file at maximum magnification to see this character. But wait! The plot thickens (or thins)! Things get even crazier.

It occurred to me that although I couldn’t paste in another smaller copy of the image there was nothing to stop me from blowing up a copy of the current large image to make it bigger, and then pasting the current large image into the even larger version.

After the flattening the image with type layer and the layers with the two smaller versions, I duplicated it. Next, I used the SI Pro 2 (Stair Interpolation Pro) Photoshop plugin from Fred Miranda Software to blow the image up. As you may know, there are quite a few Photoshop plugins that automate the process of stepping an image up without appreciable quality degradation; I’ve had good luck with SI Pro.

If you are interested in using the SI Pro 2 plugin, it costs about $25 to download from Fred Miranda Software (you’ll find the link about half way down the page on the left). I’ve used it successfully to upsize images, and also to size images precisely for printing with margin areas on particular paper sizes.

I figured I needed to blow the image up 6X to be able proportionally to fit the 18 doorways into it, for a new total of 24 regressions (I am fondly calling each doorway a “regression”, although of course technically it is no such thing).

After enlarging the duplicate six times, the new dimensions were 15,300 X 22,857 pixels at 300 dpi (or about 51 X 71 inches at 300 dpi). It still looked pretty crisp on my monitor. I was bumping up against a Photoshop limitation, namely a maximum image size of 30,000 X 30,000 pixels, and a maximum Photoshop file size of 2 Gigabytes. I was to find out about the file size limitation when I tried to save a layered version of the large-size image.

And now, a word about my hardware. I noted in an earlier story that it takes quite a bit “under the hood” for me to post-process my images, particularly when complex channel operations are involved. Specifically, I do my image processing these days on a Mac Pro with 2 X 3 Dual-Core Intel Xeon processors (four processors in total), 5 GB of blazingly fast 667 MHz DDR2 RAM, about a terrabyte of SATA disk storage, and a 30 inch Apple Cinema monitor. True, Adobe hasn’t come out yet with a native version of Photoshop, but even in simulation mode this is pretty good hardware.

On the large-sized image, things slowed to less than a crawl. Simple operations, like displaying a histogram, or adjusting a curve, took minutes. While I waited for each operation to complete, I could not only twiddle my thumbs and drink a cup of coffee, I could brew the coffee as well! A good thing I also have my Windows XP machine going for mundane tasks like email, word processing, and blog writing! Here’s to multitasking if saving a file takes half an a hour!

So now I took the 2592 X 3872 pixel image, with the 18 regressions, and pasted it in position as a layer into the 15,300 X 22,857 pixel large version. As I’ve noted, when I tried to archive a version of this image with the layers intact I hit the 2 Gigabyte file size limitation of Photoshop, but fortunately I could work on the image in Photoshop with the layers even if I couldn’t save it, and I just squeaked by just under the 2 GB limit when saving layer-flattened versions.

Besides simply elongating the corridor, an interesting optical effect happened when I combined the images. The doorways start sharp, and gradually lose apparent sharpness as the eye goes down the passage (and the image gets magnified). But at six doorways in, where the pasted original version takes over, the doorways get sharp again. This effect creates an illusion that the passage is even longer than it is, because it doesn’t seem to lose definition as the eye travels down it.

You won’t really be able to see all the doorways in the version on the web. You certainly won’t be able to see the H embedded at the end of the World without End. There simply isn’t enough resolution. To fully see my image, you’ll need to come visit me and look at it on my monitor. You could then inspect the end of the corridor at 1600% magnification and see the H.

Of course, I’d also like to make a print that is close to one-to-one. Alas, my Epson 4800 only goes up to 17 inches wide. There are a number of places around here where I can get my hands on a 9600, with a maximum width of 44 inches. A little smaller than I’d like, the print on a 9600 would end up about 40 X 60 inches. I’ll bet that with a magnifiying glass you could even see the end of the corridor.

How’s that for finding infinity’s end?

Capillarity III, photo by Harold Davis. View this image larger.

This is a close-up of a leaf, captured using my photogram technique to show the capillaries. (Here’s a similar view of a leaf from slightly further away.)

In the days of film, cross processing meant developing film with the process meant for another kind of film, like shooting Ektachrome (normally processed with the E6 process), and processing it using the C41 process (meant for Kodacolor). Film cross processing sometimes generated striking and strange color phenomena, and other times struck out entirely.

The closest digital analog to cross processing is to invert or equalize individual channels of an image in Photoshop. I post-processed this image using a number of different versions of these manipulated channels, and then reapplied the inversions to the inverted image.

To make this all easier, I’ve written a Photoshop Action that creates multiple duplicates of an image, and then seven different cross-processed versions of the original image so that I can easily see which work best.

Golden Gate Bridge from the Berkeley Pier, photo by Harold Davis. View this photograph larger.

I took this photograph during a session on the Berkeley Pier at sunset in January of 2006. This was one of the very last captures I made before the sun went down, and I remember thinking how interesting the street lights looked against the swiftly darkening sky.

When I reviewed the image on my computer I did not find it acceptable because the entire lower left quadrant had gone to black. In other words, there was far too much contrast between the light and dark areas in the photograph.

Fortunately, my post-processing skills have caught up with the way I saw the scene, and how I wanted the photograph to come out. The image you see here is the result of five different exposures from the RAW file blended together using layer masks, each layer with a different opacity, and with Normal, Screen, and Color blending mode used depending on the layer.

The five layers, each corresponding to an area in the image and each essentially representing a different exposure, are: general background (including the bridge and sky), water of the Bay including the Marin Hills, an intermediate layer for the pier fence, a layer corresponding to the darkest area of the pier, and a special layer for the street lamp lights.

For my money, the only problem with the finished image is that the layer exposed for the darkest area of the pier has a little too much color noise for my taste. Still, overall it came out pretty well.

My moral here is to hold that thought. With digital photography, if the concept behind an image is valid even if the execution is not, you should simply file the image away because there is always the chance that either the technology or your skill set will advance to a point where you can achieve the effect you wanted in the first place.

Since I’ve saved a version of the image with the multiple layers and masks in Photoshop, maybe someday I’ll figure out how to reduce that pesky color noise in the lower left!

Rainbow in the Mist, photo by Harold Davis. View this photograph larger.

I took this photograph in early October several years ago along the Mist Trail which runs beside Vernal Falls in Yosemite Valley. One is pretty sure to see multiple rainbows from the Mist Trail no matter when one climbs it, although Vernal Falls will get you wet in the spring and not in the autumn (here’s a spring photo!).

I am in the process of preparing this image for publication. When I looked at my original conversion of the photo, I felt that I’d like to see a bit more of the shadow areas, and also a bit more punch in the rainbow colors.

My new post-processing took a bit of fancy footwork. In order to capture more of the shadow detail, I imported four different exposures of the RAW original and combined them using layer masks and the Photoshop Paintbrush Tool.

A side note here: I don’t really consider this kind of post-processing High Dynamic Range (HDR) imaging. HDR creates a far great dynamic exposure range than you find in a conventional photo, usually by combining several exposures. I work within one exposure. (You couldn’t really take multiple images that are identical except for exposure of a moving waterfall and rainbow in any case.) This gives me a 5 or 6 f/stop range or possible exposures, which should be enough.

In my opinion, most HDR images look a tad artificial, and I am trying for a more natural look. (I realize that I am taking a risk in mentioning this in the context of the Rainbow image, which does look almost super real.)

My next challenge was to get more color into the rainbow. Here’s where CHOPs (short for channel operations) come into the story.

I converted the image to LAB Color mode. I duplicated the image. Next, on the duplicate image, I adjusted the A channel used Image > Adjustment >Equalize. This creates an extreme color shift in the channel. I then pasted the entire duplicated image onto a new layer in the original image, masking out everything except the rainbow. I cut the opacity of the pasted layer down to 30%, and switched the blend mode fron Normal to Color. I also added a bit of Gaussian blur to the new layer. This doesn’t impact the sharpness of the L (luminosity) channel, and the L channel is the key to managing image sharpness in LAB mode.

I repeated this process for the B channel.

The rest was just my routine workflow that I use for processing any image.

Morals of this story: It’s good to be a CHOP jock. And digital photography takes plenty of digital chops.

Capillarity I, photo by Harold Davis. View this photogram larger.

This pair of digital photograms shows capillarity, or capillar action, in a leaf. The version above has had the LAB Color channels inverted.

The behind-the-scenes tool that makes work like this possible for me is my new Mac Pro with four (count them, four!) Intel Xeon processors, and a thirty inch monitor. The monitor gives me the ability to see my images at close to 100% when I work on them, and the Mac Pro gives me the power I need to drive my image processing. This is the system that every digital photographer should have!

View this photogram larger.

Red Gaillardia, photo by Harold Davis. View this digital photogram larger.

Here are two more digital photograms of the gaillardia ‘fanfare’: one in red and one in blue. Enjoy!

View this digital photogram larger.

Gaillardia, photo by Harold Davis. View this photograph larger.

I photographed this gaillardia on a white background and, as with the bougainvillea bract, used LAB color modality in Photoshop to tinker with the hues and saturation. I think the effect is a digital-era version of a Man Ray rayograph, which are also called photograms.

Bougainvillea, photo by Harold Davis. View this image larger.

Like my Gladiolus photo, this was a pretty straight studio shot of a bougainvillea bract. In Photoshop, I converted to LAB color. I kept a copy of the original, and inverted each channel, which I applied back in various blends to the original via layers and layer masks.

Gladiolus, photo by Harold Davis. View this photograph larger.

This is a fairly straight studio shot taken with my Sigma 50mm Macro lens of gladiolas in a vase. I converted it to LAB Color in Photoshop, and then inverted the Luminosity (L) channel (by making sure only the L channel was selected in the Channels palette, and choosing Image > Adjustments > Invert). This essentially had the impact of turning the seamless white background to black without changing the other color values in the image.

Yosemite Falls in Winter, photo by Harold Davis. View this photo larger.

I captured this photograph in early spring in Yosemite Valley and archived it. It was one of those situations when the light looked glorious when I took the photo in the late afternoon from the so-called Swinging Bridge, but the capture was disappointing on my first review. (I say “so-called” because, as Julian pointed out to me with sorrow, this bridge seems pretty fixed in place, and does not really swing.)

I use Adobe Bridge for my initial screening of images, and I capture both a RAW and a JPEG version of each image. Often, the JPEG gives me a better idea of what I’ve got than the RAW version before it has been processed. The capture itself of this image in both RAW and JPEG when I viewed it in on my computer was kind of blah. The highlights appeared washed out, and the shadow areas seemed dull.

I salvaged the photo, and created an image that was more like my experience of being there than the pre-processed capture, by using multiple RAW exposures: one for the highlights and clouds, one for the shadows, and one for the mid-tones, such as the cliff. I combined these different exposures as layers, and I added layer masks set to “Hide All,” filling the layer mask with black. I then used the Paintbrush tool to “paint in” the highlight and shadow areas I wanted with varying degress of opacity (using the middle exposure as my background layer).

This is my 600th post in my photo blog, which started life modestly as a support to my Digital Field Guide, but has taken on a life of its own. If you are new to my blog, you might want to take a look at the Photoshop Techniques, San Francisco, Yosemite, and Water Drops categories.

Rio Grande near Taos, New Mexico, photo by Harold Davis. View this photograph larger.

I’ve been reading Dan Margulis’s masterful Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace. I thought I’d give some of the techniques he explains a try, starting with this image. I’m admittedly a piker with LAB color, but after this experience I’ll be using it more, and learning more about it.

You select LAB color on the Photoshop Image>Mode menu as an alternative to RGB (for computer monitors) or CMYK (for printing). You can’t really output anything in LAB, so a final step in any workflow in which you’ve converted to LAB is likely to be a conversion back to RGB or CMYK (depending on what you plan to do with the image).

The LAB color model consists of three channels. The L channel stands for luminance. Actually, this channel controls the contrast in the image, and appears in Photoshop as black and white. There’s some additional technical complexity in the way the A and B channels work, but essentially A controls the magenta to green spectrum and B controls the yellow to blue spectrum (both channels using a mechanism called the “opponent-color scheme”).

LAB color was originally specified by a standards body, the International Commission on Illumination (CIE). As the Wikipedia puts it, the LAB color model “is the most complete color model used conventionally to describe all the colors visible to the human eye.”

With the photo above, I converted to LAB color pretty early in my workflow. I was able to use the Curves dialog in Photoshop to easily color correct both the canyon area and the sky. By way of comparison, and to show what an excellent correction this is, here’s a link to a similar image from the same set that I processed a while ago without using LAB.

Moon and Golden Gate Bridge, photo by Harold Davis. View this photograph larger.

I post-processed this photograph of the Golden Gate Bridge by early moonlight from Fort Mason by sandwiching four different conversions from the RAW negative using Photoshop layer masks and the Paintbrush tool.

Recently, Reuters photographer Adnan Hajj was in the news when he was fired for manipulating a photo of Beirut being bombed by an Israeli air raid. Based on the photos published in the New York Times story, Hajj used a clone tool to crudely amplify the plumes of smoke coming from downtown Beirut. It’s not clear whether the purpose of the digital manipulation was politically based, or rather (more likely) to make the manipulated photograph more salable.

These two photographs (mine and Hajj’s picture of Beirut burning) are very different in content and purpose. Why do I bring the digital manipulation of these two wildly divergent photographs up in the same story?

Because all digital photographs are manipulated, either by software or the software’s operator. A pixel is a pixel. The processor in a digital camera manipulates the photo when it captures and saves a RAW image, and manipulates it more if it converts it to the JPEG format. And a photographer with the time, effort, and skills can “paint” a photograph pixel by pixel. Or use higher-level tools to paints pixels with greater efficiency and subtlety. As I did in my photo of the Golden Gate Bridge and Adnan Hajj did in his photo of the war in the Middle East.

So all digital photos are manipulated. What are the morals of this sea change for photographers, photo editors, and those who simply look at photos?

For photographers, just because you can manipulate an image doesn’t mean that you should do so. There’s a different duty of care and responsibility on a news photographer like Adnan Hajj, whose goal is to is to accurately portray current events, than there is on me since my goal is to produce aesthetically pleasing images. At the very least a photographer should be prepared to state whether an image has been manipulated in post-processing. Furthermore, if an image is represented as an accurate portrayal, then any manipulation should be “honest.”

I am prepared to state right now that every single photograph of mine has been manipulated. Can we all say together that digital captures straight from the camera don’t look terribly good? Unmanipulated digital photographs miss nine tenths of the potentiallity of this exciting new medium, which combines photography with digital manipulation.

Of course, if you grant that some post-processing manipulation is required for all photographs, honesty in this manipulation is a very subjective matter. This puts photo editors in a very tough position. In the New York Times article about the Adnan Hajj affair, Jonathan Klein, the chief executive at Getty Images, the world’s most important stock photo agency, is quoted as saying that the only way to avoid problems like the Hajj manipulated photos is to “employ people of integrity” and if there are abuses “not only take action, but take visible action.”

Making an example of digital malefactors is shutting the barn door after the equine creatures, digital or otherwise, have escaped. It won’t really stop the problem. In the case of the Adnan Hajj photos, the manipulation is pretty obvious at sight to anyone who is familiar with way cloning works. So photo editors need to be better educated about how software like Photoshop works, and how to spot manifestation of manipulation.

It sounds like one of the reasons that the Adnan Hajj photos got published was that the process was rushed. Slowing down and really looking is a partial solution to this issue. More important, organizations that are vouching for the verasimiltude to life of their imagery need to implement workflow that helps to guarantee the authenticity of the process.

Everyone who looks at photographs—and that is all of us—needs to know that digital photographs are artifacts that can be constructed one pixel at a time (or using tools that operate on pixels in bulk). It has always been possible to manipulate photography, but with digital photography it is necessary, easy, and sometimes very hard to detect. (Of course, the fakery is not hard to detect with the Adnan Hajj photos.)

There’s also a thin line between manipulating a news photograph in post-processing, and staging an image. There have long been charges that some of the most famous examples of war photography have been staged, going back to the heyday of silver halide photojournalism.

Skepticism is called for. The further implication is that the context of the photograph (and the person or organization vouching for it) has become increasingly important.