Big Rocks.

This photo was taken on the Cup and Saucer hiking trail east of Kagawong.

Sault Ste Marie

This is a photo of the bridge over the St. Marys river, taken from a lookout point on the campus of Lake Superior State University in the UP.

Bridal Veil Falls, Kagawong

Ok, now that my Big Annual Thing is safely underway, I should probably start posting some photos from my vacation in June. Here's one from the falls in Kagawong.

Wooden Dish

My dad won this object in some kind of game or contest at a family reunion. It was made by my mom's oldest brother, who used to be a professional carpenter before he retired. He makes things out of wood as a hobby now and gives them away. This is typical of his work.

I had to wait for a sunny day to take good photos. (I initially tried taking them indoors, but that didn't work out so well. Some of the woods he used have rather low albedo, and although my camera is significantly better in low lighting than the average consumer digicam, it does have limits.)

Summer Drawing

I've been working a bit this summer on trying to boost my limited drawing skills, and I've taken to photographing the things that I draw, so that I can go back and compare.


I was using the flip-flop for an exercise wherein I drew the same object from various angles. This was the first angle I drew it from, and then I rotated the stool and did another angle...

---

I trust everyone is familiar with the whole still life concept.

---

The exercise here was to focus on a rectangular object. I managed to find two that overlapped.

---

Here I was drawing from a photograph (that someone else took; I never saw the scene in person).

A Reflection on Weather

At this time, let us take a moment to reflect on these photographs that I took last February.


Aren't they lovely? I think they look very cool.

One of the things about a halfway decent camera...

12x optical zoom means you can actually get shots like this one, without being an animal trainer.

Dark Alley

We have, in Galion, an extensive system of alleys. Basically, these are poorly-maintained streets that run parallel to the regular ones, but in the middle of the block, between the back yards, rather than between the front yards and sidewalks. This one is always a bit dark, due to various shade-casting objects on both sides of it.

So yeah, this is what passes for a dark alley around here. Batman, eat your heart out.

This is another shot I took while experimenting with aperture settings, but this time, there's a wider range of distances between the camera and the elements in the photo, so theres' a much
more noticeable difference between F 2.5 versus F 6.4. This, if I haven't got the order I took them in mixed up, is the shot taken with the former setting. The other one is significantly less clear when you zoom in and look at the finer details.

Toothbrush Holder

This is a photo that I took while experimenting with aperture settings -- four otherwise identical photos (using a tripod) with four different settings. The shot turned out not to show off the effect very well, probably because a lot of the stuff in the shot is about the same distance from the camera as the main subject. I'll select a scene where that's not the case next time. So I cropped away the rest of the shot, and here's the toothbrush holder.

I suppose my sense of humor is a bit odd, but this toothbrush holder (which belongs to my sister, who got it because she likes turtles) has always tickled me. Clearly the design of this holder was based on the kind of toothbrushes that used to exist when I was a kid, the kind with the thin straight flat handles. But by the time this particular toothbrush holder was actually made and sold, those were a thing of the past; all the toothbrushes I've seen in the last fifteen years have contour-molded handles like the ones shown, so they don't fit. To me, this is a perfect example of product design not being revisited as it should be when the passage of time changes the circumstances in which it is to be used.

No matter; Sarah uses the toothbrush holder anyway, because she likes turtles. (And no, I didn't pose the one that's leaning against the wall. It's always like that.)

Cleaning Up Digital Photos Again

We had a holiday open house at work again, and again we took photos, and again it's my job to put them in a photo gallery on the web. I'm still working on that at work (we took a lot of photos, and I have other things to do as well), but I brought one photo home because I thought it would make for a good demonstration of some of the techniques I use to clean up mediocre photographs.


Here's a sample of what a section of the original photo looks like. This is how it came out of the camera. (Click the image to see a full-size version. Notice how grainy it is.) As I've said before, the ideal thing would be to take better photographs in the first place. A camera that just does better in low light would be good, for instance. In this case, it also might have helped to take the photo in a different part of the building, but I wanted a photo of the punch stand, and it was set up in just about the darkest part of the library, barring a closet. (The building is an old Carnegie building, and this is the part they barely touched when they remodeled in 1990, because it's near the beautiful historical dome. So no lights were added. It looks fine when you're standing there, but it's not so good for photographs.)

Also, I might have gotten slightly better results if I'd moved closer and used less zoom, but I had to take the shot when I could get it, because there were cookies on the table, and most of the time I'd have gotten shots of several people's backs. So this is the photo I've got to work with. Oh, here's a scaled-down version of the whole thing. This is, again, mostly just like it came out of the camera, except for being scaled down.

The image is a bit dark, so naturally the first thing I did was to take the levels tool and lighten it up. I discussed this step in more detail before, so I won't belabor it again now. The result looked like this.

The image is easier to see now, but if we stop here we've almost done as much harm as good, since the poor quality of the image is also now entirely too easy to see. The image is blotchy with apparently random blobs of strong color -- green, blue, red, purple, ... ick. Scaling down will help (click the image to see how much worse the full-size version is), but I'd like to clean it up a little if possible before I scale it down. So I thought maybe I'd try a despeckle filter. Despeckle is designed for much milder damage than this, but I thought I'd see what I could make it do.

I tried adaptive and/or recursive despeckling, but on an image this bad they scarcely have any impact at all, so ultimately I had to turn off those options. This is unfortunate, because it creates blur, but when an image is this badly speckled, a little blur can actually be better than the alternative.

So this is a sample of what I managed to get out of the despeckle filter. It's a bit disappointing, and if I were a real graphics artist I could probably have tweaked the numbers a bit and got something slightly better, or known which other filter to use and how to apply it, but I'm just a network administrator trying to wear an extra hat, so this is what I managed.

I was tempted to scrap it at this point.

However, I've learned that when you're trying to clear up artifacts, like speckles, from an image, sometimes what doesn't look so good on its own can be useful in combination with the original image, by using one of the various layer blending modes. Of course I had run the despeckle filter on a separate layer that was a copy of the previous image, so by changing the blending mode and adjusting the opacity, maybe I could salvage something yet.


I tried several different blending modes, but I'm only going to show you the ones that I thought worth keeping. Here it is (on the left) with the despeckled image in overlay mode with an opacity of 65%, over the (non-despeckled) lightened image. And here it is (right) with the despeckled image in screen mode with an opacity of 33%, overtop of the version with the 65% overlay, overtop of the lightened image.

The screen and the overlay have different effects, and in this case the one was effectively muting some of the speckles out of one part of the image, and the other was de-emphasizing the speckles in other parts of the image, as well as making the image overall lighter. So I combined those effects, and at this point I did my first scale-down, to 1536x1235 (a factor of the original size), using of course the bicubic interpolation. I also cropped the image slightly, which got rid of the big white thing at the left (which is actually the edge of a rectangular pillar, which is really pink not white, but nevermind).

I hadn't noticed originally, because of the other issues that the image had, but by now it was obvious that I was going to have to do something about the man's red eyes. Generally what I do in cases like this is take the lasso tool and select the most egregiously red parts of the eyes, copy, paste, make it a separate layer, desaturate it, and then turn down the opacity until it looks as close to natural as I can get.


If you want to get elaborate you can also recolorize the eyes, especially if you know what color the person's eyes are supposed to be, but in this instance, especially at this scale, I didn't deem that step necessary. Gray is good enough here.

Finally, I brightened the image up a little using the technique I discussed before, lightened that back up a bit (using levels), noted that the effect was still too strong, turned down the opacity on that layer (letting the previous, unbrightened version show through partially) until I thought it was about as good as I was going to get, scaled it down (bicubic interpolation again) one last time to 512x412 (a factor of the previous size), and this is the result:

Okay, it's still not a terrific photo, but compare it to the original:

Lightening and Brightening: Color-Safe Bleach for Digital Photographs

I recently mentioned the technique of using an unsharp mask to brighten a dark image, and someone asked me how "sharpening" can compensate for underexposure. This post is an attempt to explain. First, I should note up front that it's better to get a good clear photo in the first place. For the example, though, I have chosen one that is, in a word, not. The end result would be a bit better if the starting photo were a bit better, but using a really poor photo makes it easier to see what is going on, so for the sake of example, that's what I'm doing. This photo was taken using a cheap old consumer-grade digital camera. Here it is in mostly undoctored form:

I say mostly undoctored because I've cropped it. The curious can see more of the context here (this is part of the second photo there), but this much will do for the sake of explanation.

When a photo is terribly dark like this, the first thing most of us want to do (assuming we don't want to just throw it out) is lighten it up. There are various ways to do that, each with advantages and disadvantages (brightness/contrast, gamma correction, channel curves, ...) but the method I usually use in this kind of situation is the levels dialog box. (I'm using Gimp here, but other photo editors presumably have something similar.)

This photo was so underexposed, a significant band of values at the bright end was entirely unused in the original. So I brought the rightmost of the three sliders left to the point where the graph tapered off into nothing. Additionally, the darkest values are over-represented (note the giant spike on the left side of the graph), so I dragged the middle slider to the left a bit in an attempt to compensate. There is, however, a limit to how far you can go with this. After a little experimentation, these are the positions I ended up with:



The color is now rather dull, especially in the background. It was dark, and we've lightened it a bit, but lightening it too much more would wash it out. Instead, at this point I opted to use the unsharp mask filter. Again, I'm using Gimp, but other photo editors (well, ones serious enough to have filters) probably have something similar. Note that I ran the radius up to 50 pixels:



With a small radius (the default in Gimp is five pixels) the unsharp mask exaggerates changes in color across short distances, which helps blurry edges to seem sharper (though, technically, they aren't actually made sharper as such). With a larger radius like this, though, the changes in color are measured, and exaggerated, over larger distances. Rather than picking out edges, this picks out whole areas of the photo and makes them seem more distinct from one another. This has an overall brightening effect, making the colors seem bolder. Again, it can be overdone, but this image really needed it. Look at the difference in the woodwork around the windows from the image above to the one below. The snowflake also now looks white instead of dingy gray.


Finally, this image, in addition to being dark, was also rather grainy. Fortunately, it was taken at a larger resolution than I needed for the web, so I was able to solve this problem simply by scaling it down to a factor of its original size. Note the use of the cubic interpolation: chintzier scaling algorithms like nearest neighbor are not recommended, as they can actually exacerbate image quality problems. Here is my result:



It's still not the best photo ever, but if you scroll back up and look at the original again, you will notice how much worse it was. This really is an improvement. Of course, starting with a better photo initially will yield better results.