Last Sunday I had the pleasure to participate in making an aerial map, a very detailed overhead image of Dolores Park taken from 200’ to 400’. The fine folks at Public Laboratories have published the result. A couple of direct links: a full page slippy map and aerial video. I also got a nice single shot of the new Dolores Park playground (full size).

This kind of mapping comes from Public Laboratory, a cool non-profit helping ordinary people make their own aerial maps. You don’t need a satellite or fancy cameras on a plane to make a “satellite map”. All you need is a kite or balloon, a cheap point and shoot camera, and a little software.

I particularly like how low tech the setup is. The lift comes from something as simple as a mylar balloon filled with helium (we used a car dealership ad balloon). No fancy cradle, just some rubber bands holding the camera to a plastic soda bottle with the soda cap as the clamp. And no special camera, just an ordinary consumer camera set to continuous drive with another rubber band holding down the button. Launch, fly for awhile, pick the best photos, then stitch them together into a map. (That last part takes some effort; you can do it by hand in Photoshop or use the MapKnitter web app).

The Dolores Park shoot was mostly for fun, a nice way to be outside on a sunny day and excite a bunch of kids. But Public Laboratories participates in more serious projects too, like documenting the Deepwater Horizon oil spill. A key thing about this kind of mapping is anyone can do it relatively cheaply and quickly. You can also fly other sensors, they’re doing a lot of work in environment monitoring,

Many thanks to Stewart Long and Bobby Sudekum for letting me tag along on the fun photo shoot.

Photoshop CS2 is the most expensive piece of software I own. I wanted to migrate my license to Vista and it wasn’t as simple as I’d like. The main challenge is CS2 is an old version and Adobe claims it does not work on Vista. Here’s how to make it work anyway. I’m assuming you have a legal copy: do not steal software.

1. Get install media. If you have CDs you’re in business. Otherwise write to Adobe support and tell them you need to download a CS2 installer. Do not mention Vista; they’ll tell you that you have to spend $200 on an upgrade to CS3. They sent me an FTP link with a password for Photoshop_CS2_9.0_WIN_ACT.zip.
2. Unzip the install media. Don’t just double click the installer and run from the virtual folder; the Adobe installer will break with the meaningless errors .\AutoPlay\main.ini MAIN_FILE_VERSION PRODUCT_REGISTRY_PARENT PRODUCT_REGISTRY_KEY. Extract files and run from the new folder.
3. Run the installer. Grant it permissions via the usual Vista UAC popup.
4. Run Photoshop. Activate it.
5. Allow Photoshop to update itself. Grant the Updater permissions via UAC.
6. Reboot. (Screw you, Adobe).
7. On reboot, you’ll get an error that Vista blocked programs at startup because they asked for permissions. Run the Vista tool recommended in the error box and disable all of Adobe’s startup tasks (gamma control, updater). The error will go away and your computer will boot faster, too.
8. Enjoy Photoshop!

I’m sure there’s some lurking Photoshop / Vista compatibility problems. But I was able to edit a raw photo from my camera, that’s all I really care about. I didn’t test color management and printing, I could see potential problems there.

I finally figured out why the colours in all my photos on Flickr look wrong: Firefox is stupid and broken and doesn’t honor embedded ICC color profiles. I’ve been doing all my photo editing in Adobe RGB and saving Adobe RGB JPGs with ICC profiles attached. Firefox ignores the profile, interprets the images as if they’re sRGB and the colours end up all washed out.

One solution for this is to tell Photoshop to just do all the work in sRGB, the default that color-stupid apps like Firefox assumes. But I’m stubborn and want to use Adobe RGB when editing on the off-chance I want to print something. The solution is to convert to sRGB right before saving the JPG, via “Edit / Convert to Profile”. In theory the Photoshop “Save for Web” option will take care of this for you, but in didn’t seem to work for me and save for web does other things wrong anyway.

Awhile back I bought a ColorVision Spyder to calibrate my monitors. It works great, makes a huge difference. Now that I understand better how profiles work (or don’t work, in the case of Firefox) I think I can get consistent colour all the way to the printer.

Flickr started as a side project from a high concept online game that was really an excuse to socialize and collect things. The photo thing took and the Flickr guys were smart enough to cultivate a success when they found it, but games still pop up inside Flickr from time to time. Users are the ones inventing Flickr games now. I just noticed a new one, a variety of “your photo is good” badges people put in the comments of photos like this one. Flickr Gold lets you give out gold medals after you get one, Flickr Poker and High Five let you hand out several awards a day and accrue kudos to photos with lots of awards. Sort of a carebear version of an old Flickr game, deleteme.

These games seem a bit silly, but I think people really like this kind of tokenization. It’s not enough to say “this photo is good”, the ritual of collection and the structure of game rules makes it seem more significant for people.

On Flickr there’s a useful thread describing a quick trick to make colours pop. You take your photo, convert to Lab color mode, then compress the a and b colour channels. This has the effect of making colours generally brighter at the cost of destroying some detail in intensely coloured parts of an image.

Not always a good idea, but it can make certain pictures look better. I’ve done it to about 5% of my photos I put on Flickr. To make life easier I’ve created a Lab color pop action set for you to download. First you should try the method manually once to understand what it’s doing.

Thanks to Flickr it’s finally easy to use geotagged photos. Alas, the workflow is still very clumsy for using a GPS to geotag. Here’s what I came up with for Windows.

1. Get a GPS. Six months ago the Garmin Forerunner 301 was the smallest thing I could find with a USB interface. The newer Forerunner 205 looks good, as does this upcoming Sony thing.
2. Walk around with the GPS tracking your location and take some photos. Make sure the clocks on the GPS and the camera are synchronized.
3. Dump the GPS track log to a .gpx file. Garmin drivers are here. I couldn’t find Garmin software to just write a GPX file, but get301tracks works. Or you can use RoboGeo to download and save GPX.
4. Run RoboGeo ($35). Follow the instructions for “3. From the tracklog in a GPX file”. Pick your JPGs, then pick your GPX file, then click “Write the location data to the EXIF headers” to create new geotagged JPGs.
5. Tell Flickr to Import EXIF location data. It’s a one time preference setting.
6. Upload your photos to Flickr via your usual method.

There’s a much easier way to do all this inside RoboGeo. Just plug in your Garmin GPS, point it at your photos, and upload to Flickr. No messing with EXIF tags or GPX files or anything. But while that’s admirable for its simplicitly it’s totally the wrong thing for serious image management. You want the raw location on your hard drive so five years from now you can use it again.

The flaw in the above workflow is it doesn’t put the geotag data in your raw image originals. RoboGeo supports DNG but not CR2. Ugh. Keep the GPX files around.

Congratulations to the Flickr team for their new geotagging interface. I think this is the first time it’s truly easy for normal people to indicate where they took their photographs in a way others can benefit from.

It’s a bit late for a pony request, but I sure wish geocoding interfaces had a notion of uncertainty along with location. Ie: latitude, longitude, and radius. A photo geocoded to “San Francisco, CA” is different from a photo geocoded to “659 Merchant St. San Francisco, CA”. The former has a radius of about four miles. The latter, 20 feet.

Tagging your photos via a map interface gives a natural radius: a few pixels’ width in the map the user clicked on. GPS devices have a natural uncertainy measurement, too. So the data’s available, why not store it?

I have about 300 photos from Australia I intend to edit and put online before I leave for France. But as anyone who takes a lot of digital photos will tell you, workflow is still a huge problem. Here’s what I do now:

• Stick flash card in a usb device, copy photos as files
• Use Photoshop to convert raw to JPG previews
• Sort through jpegs in ExifPro to find the good ones
• Symlink (by hand) the raw files for the good photos
• Lightly edit raw files in Photoshop, correcting exposure and cropping
• Upload edited photos to Flickr

None of the automatic tools for photo workflow work for me. Either the image edit tools are bad, the database features are too limited, the software is too slow, or the UI is awful. And all of the fancy tools want to store my data in some proprietary database where I can never get at it. The workflow above is manual, but at least the data is transparent.

I was curious about how many tags people put on their photos in Flickr and whether more tags help you get into interestingness. So using their fabulous api I wrote a quick program to count the number of tags for photos in interestingness and compare them to photos in the huge FlickrCentral group. Numbers are averages and standard deviations for both the author’s own tags and for tags volunteered by others.

               Author tags  Viewer tags   
FlickrCentral  7.5 (9.1)    0.1 (0.4)
Interesting    8.0 (6.4)    0.7 (2.2)

What’s it all mean? Not much. The average FlickrCentral photo has 7.5 tags and the average Interesting photo has 8.0 tags. Well within the standard deviations, not a good predictor. Photos in Interestingness do have measurably more viewer contributed tags but that’s not a surprise; photos in Interestingness have more viewers.

So not an interesting result, but I love how easy it is to do this kind of data mining with Flickr.

(For the record, I compared the top 100 interesting photos for each day so far in July, 1900 photos, with the 500 most recent photos in FlickrCentral).

A lot of my photos from Australia have bright blue skies with dark trees in front of them. Like this. One thing that irritates me when editing them is the tree branches have noticeable blue tints.

The problem is the branches are only a couple of pixels wide. And digital cameras aren’t true colour sensors, but rather a Bayer filter mosaic of red, green, and blue sensors. So there’s some aliasing in the mosaic pattern in the colour sensor. The raw convertor has to be particularly clever to figure out the true colours in the image. To put it another way, a digital camera sensor is going to have a very hard time with colour features only a couple of pixels wide.

Here’s a comparison of four raw convertors: Canon’s Digital Photo Professional, Rawshooter Essentials, Adobe Camera Raw, and DxO Optics. To keep things consistent I used full defaults for each convertor. So ignore the exposure and overall colour, just look for bleed. To my eye the DxO image has the least bleed. The blow-up doesn’t look as sharp in DxO, but I think that’s actually more faithful to the reality.

Tinkering around I think I can get the best-looking image out of DxO. It has lots of fancy camera body and lens-specific corrections and, apparently, a smarter demosaicing algorithm. But this kind of pixel-peeping doesn’t make me a better photographer. And DxO is a pain in the ass. So I’ll just stick with Photoshop. Still, the technological differences are interesting.

My blog post about HDR photos and Photoshop has been unexpectedly popular. I wish I’d done better explaining HDR, but I’m still working on understanding it myself.

In addition to Photoshop there are a lot of other HDR tools out there. Photomatix is one really easy to use standalone tool with a free trial for Mac and Windows. It seems a bit more fast-and-loose than Photoshop’s nerdy careful approach, well suited for exploration. I particularly like the “Exposure Blending” tool where you can throw a bunch of photos into it and quickly come up with a single tone-mapped JPG. If you want to play with HDR and Photoshop seems too difficult, check out Photomatix.

I’ve updated my HDR Flickr set to include 3 different HDR processes on the same set of pictures: Photoshop, Photomatix automatic, and Photomatix tone mapping. Amazing what variation there is.

I’ve been playing with my DSLR for months now without understanding what shutter speeds and apertures really do. After doing a bunch of reading on the science of exposure and the intuition of exposure, I finally get it. At least I think I do and now I’m going to embarass myself by revealing my ignorance. Please tell me if I get something really wrong.

A photograph is properly exposed if the camera has the right aperture, shutter speed, and ISO film/sensor sensitivity to capture most of the light coming in from the scene. The actual science is complicated (the magic word is EV), but the way photographers talk about it is in terms of “stops”. If a scene is twice as bright then you need one less stop in the camera to make your photo. Conversely, opening up a stop doubles the amount of light recorded.

Shutter speed is one way to adjust stops; if you double the length of the exposure you double the amount of light captured. Common hand-held exposures in one stop increments are 1/30s, 1/60s, 1/125s, 1/250s, 1/500s. My Canon 350d lets me adjust shutter in 1/3 stop increments such as 1/30s, 1/40s, 1/50s, 1/60s. So I have to click the wheel three times to double the amount of light recorded.

Aperture (or, confusingly, “f-stop”) is another way to adjust exposure. Opening the lens aperture one stop will double the amount of light coming in. The numbering is a bit confusing because of the geometry; doubling the f-stop number will cut the amount of light coming in by one quarter, two stops. Common one stop increments for aperture are f2, f2.8, f4, f5.6, f8, f11, f16. My Canon 350d lets me adjust aperture in 1/3 stop increments such as f2.8, f3.2, f3.5, f4.0. So just like when adjusting shutter speed, I have to click the wheel three times to double the amount of light recorded.

Finally, the third way to adjust exposure in a camera is to change the ISO sensitivity of the digital sensor (or film). Doubling the ISO will make the sensor twice as sensitive to light, which is about the same as doubling the amount of light recorded. Common ISOs in digital cameras are 100, 200, 400, 800, 1600. My 350d only lets me adjust ISO in full stop increments (not 1/3), so every time I “click the wheel” (or push the button) I’m doubling the amount of light recorded.

That’s a lot of detail, but what it boils down to is:

• 1 stop = twice the light recorded
• Doubling exposure time = 1 stop
• Halving f-stop number = 2 stops
• Doubling ISO = 1 stop

A modern camera with light metering means you can mostly be ignorant of this stuff, just put the camera in P mode and click the wheel to trade off shutter speed and aperture for artistic effect. But now I understand just what I get by bumping up one ISO level: I can cut the exposure time in half. And this information is a foundation for more complex topics. The next thing I’m trying to understand is dynamic range; just how wide a range of EV various scenes have, how many EV levels my camera can capture in a single photo, and how many EV I can print or view on a screen. This article is a good start.

Photographs in extreme lighting are a compromise. Either the highlights are overexposed or the shadows are underexposed. But with HDR techniques you can take multiple photos at different exposures, then combine them into a normal image with fine local control over exposure. It’s a bit like the Shadows/Highlights adjustment, only with real data instead of fakery.

Photoshop CS2 makes HDR easy for normal people. Some intro material on HDR can be found in this tutorial and this article. The HDR Flickr Group is fun, too.

I made a small HDR experiment while in Mendocino. The result isn’t great; boring composition, the fence looks unnatural, and I didn’t have enough range of source images. But it was a good exercise. Here’s how to do HDR in Photoshop CS2:

1. Take several photos of the same scene with a tripod at a range of exposures. No individual image will be right, but you’ll get details in both shadows and highlights.
2. In Photoshop CS2 select File / Automate / Merge to HDR. Choose your images; RAW is preferred. Unless your tripod is perfect you want “Attempt to Automatically Align Source Imges”.
3. In the preview, just click “OK”.
4. Choose Image / Mode / 8-bit to bring the HDR down to a normal image.
5. You have several options for methods for squashing the image. “Local Adaptation” is the powerful one. You can also use the toning curve like an uber-levels tool to tune the exposure.

This is only the simplest possible conversion; there’s a lot of art here in manipulating the toning curve, choosing your exposures, etc. I’ve also found it’s useful to convert to 16 bit first to do some conventional exposure editing, then convert to 8. But still, this is enough to get started.

Adobe Camera Raw is a good RAW image convertor that’s part of Photoshop. I spent a lot of time trying to figure out how to batch convert a whole directory of photos, with confusing “actions” and “droplets”. Then I learned it’s a lot simpler than that.

1. In Photoshop, choose File / Open.
2. In the open dialog select all your RAW files.
3. Click “open”. Don’t panic; it won’t chew up your RAM.
4. In the filmstrip on the left select all photos.
5. Choose your image settings. If you’re converting a lot of images, you probably want “Camera Raw Defaults”.
6. Click the “Save N Images...” button in the lower right.
7. Select format: JPEG. Adjust whatever else you want; file naming, JPEG quality, etc.
8. Click “Save”.
9. Click “Done” and wait patiently.

I’m doing this in Photoshop CS2, but I suspect it works in Elements just as well since really it’s the standalone RAW tool that’s doing the work.

Bonus tip: Microsoft has a RAW Image Thumbnailer PowerToy that lets the file browser make thumbnails of your RAW images, just like JPGs. Works great.

Qoop is a clever service that prints books of your photos. It has amazing integration with Flickr. Go to Qoop, look at your Flickr photos and choose which to print, enter a credit card and a few days later a nicely bound book of pictures comes to your house. Couldn’t be simpler.

I ordered a book of 168 of my photos from Switzerland. The 8x10″ book format is great; easy to flip through and look at pictures, easy to put on a shelf to preserve. Qoop also offers posters and calendars and soon, postcards.

The print quality is a compromise. The perfect binding and paper quality are terrific. But it’s not a photo printer, it’s more like a high quality color laser printer. The color matching isn’t great and sometimes the ink is a bit uneven. But the books are relatively inexpensive (my order came to 40 cents a photo) and a heck of a lot nicer than printing a bunch of 5x7″ and pasting them in an album. Their customer service is very friendly and responsive, too. I’ll definitely be ordering more books.

I’m amazed at how convenient the process was. I assume the core enabler here is the Flickr API which allows Qoop to get at my photos. More web services like this, please!

I like my photos taken with the cheap Canon 50/1.8 prime lens (like these) more than the ones taken with my Canon 24-85 zoom lens (like these). Prime lenses really are sharper and have clearer colours than zoom lenses. But zooms are convenient, particularly for walking around.

The usual suggestion for an excellent all-around zoom is the Canon 24-70 f2.8L. But it’s very expensive and at 950 grams, too heavy. An alternative is the Tamron 28-75 f2.8. It’s got the same basic specs as the Canon, but a third the price and half the weight. Folks seem to love this lens.

So I bought the Tamron and did a quick comparison with my Canon 24-85. The photos are ISO 100, f/8, shot on a tripod, JPEG straight from the camera. In the sample above two things stand out. The colour on the head is better in the Canon, but I think that’s just luck of the clouds. And the reflection in the candlestick on the left is so much sharper in the Tamron! I think that’s for real. We’ll see.

This comparison is a bit weak; it’s hard to really measure lenses. But I’ve got an example that convinces me this lens will work out. See the Flickr set for all of the photos.

Canon has a yearly amazing rebate program; if you buy three qualifying items you get a 10-30% rebate on all of them. Rebates cover many things (including camera bodies), but you have to buy things by January 15. Here’s a list of all the prime lenses that are covered by the program. Prices listed are the cheapest at Amazon, rebate is the maximum 3x rebate.

Length	Aperture	Weight	Retail	Rebated
20	f/2.8	405g	$420	$375
28	f/1.8	310g	$400	$355
50	f/1.4	290g	$315	$270
85	f/1.8	425g	$340	$295
100	f/2.0	460g	$390	$345
135	f/2.0L	750g	$890	$785
200	f/2.8L	765g	$660	$585
300	f/4L	1190g	$1150	$1015
400	f/5.6L	1250g	$1100	$980

All lenses have ring USM autofocus except the 50 which is micro USM. The 300mm is also image stabilized. More data on the Canon lens chart PDF.

I love my new Canon 350D. It’s the first SLR I’ve owned, and the pictures are automatically 10x better than my little Canon Elph S400. It’s the only time in my life where spending money made me a better artist.

But inexpensive digital SLRs don’t shoot a full 35mm frame. Instead they have smaller sensors. The 350D has a “1.6x crop factor”, which basically means your pictures look OK but your lenses are more telephoto than you’d think. A 35mm lens acts like a 56mm lens. If you want a wide angle 28mm shot you have to buy a 17mm lens. Really wide angle lenses for these small sensor cameras are very dear.

Canon has tried to turn this liability into an advantage by selling “EF-S” lenses, special lenses that only work on small sensor cameras. They go deeper into the camera body, and in theory allow for cheaper / lighter lenses. All I know is they don’t work on all cameras, just these small sensor ones.

And while right now most digital SLRs are small sensor, there’s reason to think that Canon may soon switch to all full size sensors. The newly announced full sensor Canon 5D is a bellweather here. At $3300 it’s a lot more than the $800 for a 350D, but a lot cheaper than the $7600 that the current cheapest full sensor camera costs. Folks seem to be excited by the 5D full frame. Wide angles that work like wide angles, better viewfinder, what’s not to like?

All of this is a long-winded way of suggesting that EF-S lenses may not be such a good investment, that in four years you may not be able to buy a good camera that uses them. Which is sort of an awkward position for Canon. They’ve made some serious EF-S lenses, will they abandon their customers?

My camera doesn’t have raw output, but it does have three JPG quality settings. I just tested empirically to see how they correspond to the JPEG Q parameter:

Setting	Size	jpg Q
Superfine	1824k	98
Fine	1148k	94
Normal	580k	81

Given that quality of 75 is considered adequate in most JPEG software, there’s some comfortable headroom here. Fine’s 94 quality seems plenty for the photos to me and saves almost half the space. Then again with 1 gig of Compactflash for $83, there’s not much need to economize storage. Bandwidth is still expensive, I’ll have to remember to recompress the images before uploading to Flickr.

CCD sensors used in digital cameras can’t sense colour, only brightness. So how do you get a colour image? Typically by putting little red, green, and blue filters in front of the sensors on the CCD like the GRGB Bayer pattern depicted here. The camera software then has to do some funky tricks to compose a proper colour image.

Foveon has a sensor technology called X3 that works around this problem by stacking several translucent sensors in a single pixel. In theory this should give sharper pictures. So far the Sigma SD9 is the only camera out there with the Foveon sensor ($1500). Alas, the reviews suggest this first camera isn’t quite up to the sensor.

Canon’s photo software loses data. Their tool autorotates images and renames files sensibly. But it blows away a bunch of the metadata from the photo! Better to copy via Windows. If their software rotates an image for you, you lose at least the following EXIF tags:

I wrote my own tool to read the EXIF data in the image and rename and timestamp files appropriately. What a pain in the neck. EXIF isn’t a good standard; I decided exiftags does the best job of reading EXIF files, including Canon’s extensions. Converting timezones is a pain.

I was going to autorotate the images but as I note below the software that claims to do this losslessly doesn’t.

Maybe I’m obsessive? Well, now I have a Python library.

jpegtran claims to losslessly rotate and crop JPG images. Neat trick - instead of modifying and recompressing the image, jpegtran just fiddles the vectors.

Only it doesn’t work, at least according to the Gimp. Above is the difference between the original image and the one passed through jpegtran. I’ve boosted the gain a lot - the difference histogram is the image on the upper right. The differences had pixels as much as 17/255 white! It should be all black.

Is this a flaw inherent in jpegtran? Is it an artifact of the Gimp’s jpeg decoder? It’s not just jpegtran; Canon’s own software has the same artifacts. And I don’t think it’s just the Gimp; I get the same results with djpeg and pbmplus.

I’ve been using ofoto for printing images from my camera. They do an OK job but they overprocess - I see unsharp mask damage. I’m ready to try another printer.

I haven’t found a good review site, but looking around I did find PrintRoom.com. Their prices are better than ofoto, you can tell them not to do any enhancements, they have a FAQ for pro photographers, and they even provide an ICC profile if you want to colour match yourself. Worth a shot.

Unfortunately WinGimp doesn’t have any ICC support. Raph has a good page explaining why: it’s hard and there are patents in the way.

If you’re hardcore about evaluating digital cameras check out the ISO 12233 Standard Test Image, produced by I3A TC42 WG18. A finely detailed monochrome test image with features as small as .1mm. Take a snapshot of the $150 ISO image, measure your spatial frequency response, and you know how good your camera is.

I learned about ISO 12233 on Imaging Resource, as part of their fanatically detailed page about camera testing. The review of the S400 has the 2272x1704 WG18 ISO 12233 image if you need. Also, the viewfinder test confirms I’m not crazy - the optical viewfinder really is way off center.

I’ve had my new digital camera, a Canon Elph S400, for a week or so now. I’m really impressed with it, particularly the quality of the colour even on full automatic settings.

I’ve had a lot of fun with the macro setting. Cliche flower photos, but they’re fun to do.