Vroman’s Nose

Touch the earth, love the earth, honour the earth, her plains, her valleys, her hills, and her seas; rest your spirit in her solitary places. -Ernest Dimnet

Had a small adventure on Sunday, a short hike up Vroman’s Nose in Middleburgh. I never even knew this place existed, but it’s a pretty amazing bit of glacial geology, that is if you find words like escarpment amazing.

The Nose is a site of historical Revolutionary War significance and probably has one of the highest ease of hike to scenic view ratios anywhere, ever. Take the green path and you have a casual walk of about a mile ending at the top of the escarpment, with a drop of about 500 feet below you. There are no railings, and Laura made me nervous more than once by sitting on and leaning over the edge.

I’m not afraid of heights, I’m just afraid of falling.

So you get this amazing view of the Schoharie valley, but the real attraction is the nesting pair of peregrine falcons, who have already fledged at least one young. Go there and you can see why they like the place; from the cliff face, they have acres and acres of fields to hunt in. Peregrines are the fastest bird in the world; they can go 200 mph while divebombing prey that they can see from two miles away. They’re so badass that they usually hunt other birds that they actually catch in flight. What’s so amazing about a successful nesting pair is that there are only two to three thousand such pair in the U.S., their decline due to the use of DDT weakening eggshells, just like the bald eagle.

There’s your education for the day.

I was able to get a couple shots of one of them who flew in and perched on the cliff to keep an eye on us. I had the 75-300mm telephoto zoom, but this is not a quality lens; I was happy to get the snapshots I could but there are some optical issues with the photos bothering me. The 70-200mm f/4L is in my future somewhere. We also saw a small flock of cedar waxwings, they just kind of hung out and let us take a bunch of pictures.

Check it out. Next weekend, hopefully we can get up to the Adirondacks for some more adventure, and actual snow. There’s still no snow in Albany, WTF.

So this is a photo of a girl I’m extremely fond of…

I am extremely fond of her.

That’s all for today.

Almost Foodies Soup Swap

Now that I have the workflow issues mostly sorted out, I can return to blogging as usual. On the 23rd Almost Foodies held their annual soup swap and were kind enough to invite me, against everyone’s better judgement. I brought the party, and a ginger squash soup I’ve been practicing on. And my camera.

It was really the best time you could ever have with soup. Bring six quarts, go home with six quarts; I already finished three, one of them being a huevos flamencos, minus the heuvos, inspired by the same dish at a little place called Carmens in Troy. Now, it’s really good at Carmens, but this was easily on par if not better (once the huevos are added).

I continue to enjoy soup; you get to enjoy photos. Sorry, life isn’t fair. Protip: if you click through flickr, there’s a rare glimpse of yours truly. That’s more of a warning than a promise.

Open Source Noise Reduction Comparison

Alright. After a few days of searching, trial and error, and teeth-grinding frustration, I’ve got some answers about noise reduction with RAW camera files in GIMP. If you don’t want to look at a picture and see lots of numbers, here’s the summary:

Use whatever you like.

Yes really. The differences between available methods are academic and only truly come into play in extreme circumstances — and if you’re denoising to the maxxxx! you should probably just trash your photo and start from scratch. If you want to hear some more specifics and get some other thoughts, keep reading.

Important note: I’m using a Canon 300D and an ancient computer (2.0GHz, 1GB RAM, GT6600 GPU, Win XP). This imposes some restrictions; I can’t just shrink a photo to help hide noise because I’m only working with 6.3MP to start with, I’m dealing with massive amounts of noise even relatively low ISO settings, and using anything too system-intensive takes a long, long time. Basically, I need something that actually works, and with a light footprint.

Here are four versions of a 100% crop of the same photo I’ve been using. You can see where there are some fine details and some expanses of color that should be more or less smooth, basically the two extremes we’re working with when denoising. Ideally I’d have some darker areas but that might be for a future test.

This is output straight from the camera, and denoised in UFRaw, G’MIC (using the GREYCstoration anisotropic smoothing), and the Wavelet Denoise plugin. This was taken at ISO 400, f/3.2, 1/4000 sec, with a 50mm prime lens.

The camera RAW crop speaks for itself. Plenty of noise in the out of focus areas, and really ugly noise at that. This is what we need to fix.

UFRaw has built in noise reduction, controlled with a single amount slider. It pays to have a light hand with this slider; here I set it to 60. This did a good job of smoothing out the noise but at a significant loss of detail. Moreover, there’s an overall “soft” quality that I don’t quite like — what I described in an earlier post as everything looking like it was made of wet jello.

G’MIC is extremely powerful. The noise reduction is basically the old GREYCstoration tools bundled in, along with a tons of other really useful stuff. It’s worth installing just to have more and more powerful toys in GIMP. There are two problems with using it to denoise though: it’s complicated, and it’s really slow. As in, for me, it takes about two minutes to render a 6MP file slow. And here are the settings I used:

Amplitude: 40.00
Sharpness: 1.00
Anisotropy: 0.30
Gradient Smoothness: 0.60
Tensor Smoothness: 1.10
Spatial Precision: 0.80
Angular Precision: 30.0
Value Precision: 2.00
Interpolation: Nearest Neighbor
Fast Approximation: ON
Iterations: 1
Channels: All
Tiles: 1

That’s a lot of parameters. But the result is not bad, details are preserved and noise is obviously suppressed, with a few artifacts. Now, I could probably spend some time figuring out how this all works and get a much better result. The problem is, that’s going to take a lot of time. And with waiting two minutes to see each result, and that many options, I think it’s safe to say that it’s not really worth it. G’MIC’s noise reduction is very powerful, probably the best open source solution in terms of quality, but it’s just too much. Sorry.

Wavelet Denoise. This plugin’s complexity is somewhere between UFRaw and G’MIC. Well, anything is more complex than UFRaw’s single slider. But Wavelet Denoise gives you the option to work in YCbCr or RGB space, tweaking each channel for amount and level of detail. I used the following:

Y: 500
Cb: 1500
Cr: 1500
Detail 1.0 (all channels)

This is a fairly quick and dirty adjustment, and the results are also dirty but good. We have more sharpness than UFRaw, fewer obvious artifacts than G’MIC, and acceptable noise removal. And bonus, because there are fewer confusing options than G’MIC, I’m confident that with practice I’ll be able to refine these results. The best part is that this renders in about 15-20 seconds for me, or about the amount of time as an Unsharp Mask. This is much easier to incorporate into my workflow than adding two minutes to every photo I want to process.

So.

I stand by my summary: use whatever you like. Right now, I happen to like Wavelet Denoise because it offers a fair amount of control and returns good results. UFRaw would be fine if I didn’t mind the hit in sharpness and wanted something extremely simple and convenient. G’MIC is the backhoe for your underground pool: the best tool for the job, if you have a huge job and know how to use it.

Later I’ll run down more specifics for using Wavelet Denoise.

Edit: It’s been pointed out that I’ve been using an older version of Wavelet Denoise; a couple weeks ago a new Windows version was compiled, the main difference being that it now handles CIELAB color space. I’ll grab that and do some more work, and later post an entry with details on using Wavelet Denoise specifically. Thanks to everyone who’s given me input so far, and I hope that those of you who found this through searches for open source noise reduction solutions are finding it useful.

Wavelet Denoise plugin for GIMP results

The results are in.

For now.

Wavelet Denoise seems to be the way to go. Compare the Wavelet Denoised version with the version straight from UFRaw. You’ll see it right away, but especially look at the hood ornament. Only a very small loss of detail in the sharp, in focus areas, and much cleaner swathes of out of focus color.

I did slightly reduce the exposure, which does help, but this difference is dramatic. Even better, it’s quick rendering in about 15-20 seconds. Much, much faster than G’MIC.

When I get a chance I’ll fine tune some more and do a quick runthrough with 100% crop examples of straight RAW, UFRaw’s noise reduction, G’MIC, and Wavelet Denoise. For now, I’m pretty happy, but I do have a lot of work ahead of me.

Earplugs (i.e., less noise)

Installed the G’MIC plugin for GIMP last night, which includes the GREYCstoration library to try some more denoising options. Very powerful. Very slow. But it looks like it will work, once I get all the settings dialed in. As soon as I get some time to play, I’ll toss up some more photos.

I may also try the wavelet denoise plugin but I don’t think it’s much different from what UFRaw does already. If anyone has any input (that doesn’t include the words “pirate” and “noise ninja”), I’d love to hear it.

Noise and UFRaw and a Bentley

Processing RAW files from the 300D is proving to be frustrating; for a few seconds last night I considered just tossing everything out the window. I didn’t, because then I’d have to go in the back yard and pick everything up anyway, and fix a window.

So, this picture of a Bentley that someone parked in Troy (really? really?) was shot at ISO 400, and is a good example of what way too much ugly noise looks like. You can only see hints of it at this size; click and look at the larger sizes and then feel bad for me. The good news is that this is with no noise reduction in UFRaw, so I have somewhere to go. The bad news is that enabling any noise reduction in UFRaw makes it look like everything’s made of wet jello.

The solution may be a light hand with noise reduction in UFRaw and them some sharpening in GIMP. I am happy with the amount of control I have with RAW images, just not happy that the end result isn’t up to par with JPG files from the camera. Due to user error. But still.

Oh, so this photo, of a Bentley, I’m guessing a late-80’s Continental but someone correct me, parallel parked on River Street, that was fun to take.

How to improve your photography

Keep doing it.

No, seriously, new photos soon, I promise. Switched to a RAW workflow (finally) using UFRaw and GIMP, getting used to that. In the meantime, I guess read my Twitter or something.

How to make your photos more interesting

Damned if I know.

More from Troy

Winter is a tough time for shooting. Less daylight, the cold, fewer local events, it all makes me very unmotivated, I just want to sit in my room and listen to Silversun Pickups all day. I have a few things to work on but it’s just a slow season. A little color from last saturday’s Troy Farmers Market and walk around the city helps though…

Yarn at the Farmers Market

We have a rule: you see a chalkboard sign, you shoot it

A geometric look at Troy’s Atrium

Now looking into a k-mount adapter for the Canon, DIY lightboxes, and finally working in RAW with GIMP.