Bart's Notebook - Photography

Bigfoot in Grand Canyon, circa 1969

2010-10-25T21:15:00.000-07:00

I've been scanning in mass quantities of slides my Dad shot when I was a wee lad. I was working through some shots from a family vacation to the Grand Canyon and I couldn't help but notice what looks like Bigfoot strolling up the hillside in that trademark hunched gate we've all seen in the various blurry Bigfoot shots. It's just blurry enough that I can't quite figure out what that thing really is and there are no other similarly textured objects in the scene that I've found. If I had to guess, I'd say it's the snarled remains of a Juniper tree--the texture looks about like Juniper bark.

Some technical info: This slide is a standard 35mm slides (I don't know what type of film). I scanned it at 4800dpi on my Epson V500. I've chosen this resolution because it's high enough to capture the film grain and so if I were to ever perform noise reduction, the grain could be effectively removed. Scanning at a a lower resolution will still capture the image detail, but could cause aliasing of the grain thus making it harder (or impossible) to remove. I've sharpened the inset to try to clear it up. The black smudges seen on the inset (there's one on bigfoot's shoulder and another about halfway between "him" and the right edge of the photo) are a defect in the slide itself I've been seeing on most of these old slides. I don't know if it's just something that happens with aging or if it was there on the original. I had the sharpening shut off for my scanner, but I suppose it's possible it's the cause. I'm intending to have a look at the slide under a microscope to see if the artifact is really on the slide.

Bart

Morning Frost

2009-12-12T13:13:00.001-08:00

I was rushing out to work and getting ready to scrape off a bunch of frost that had formed on my truck (it was about 10 F out) and noticed the frost crystals were more interesting than usual. So I put down my bag and ran in to grab my camera. Used the 50-200 zoom lens with the Raynox 250 attached and my hotshoe flash (handheld). Some of the frost was on the truck window so I could hold the flash behind the frost to back-illuminate it.

Also crawled around on my lawn and got a couple shots of frost on dead leaves in the grass.

You can visit the gallery here:

2009.12.10 Frost

Japanese Garden

2009-10-17T22:45:00.000-07:00

I haven't been getting out to shoot much at all for months--been very busy at home and at work. So I finally snuck out Saturday morning to the Japanese Gardens to get some of the fall colors.

Used a flash here--resting on the ground to my left

HDR comp using Photomatix. Replaced some burned out sky with a forest background (that's the green you see through the red leaves). Multitude of other photoshop retouching.

Simple shot of leaves at the entrance to the garden.

Fixing a Tight Lens Hood

2009-01-20T20:11:00.000-08:00

Ever get a lens hood that is either too tight or it sticks? I've had two such situations. One lens hood got a bit sticky after a lot of use at the beach--I think some sand got into the hood groove and might've scuffed it a bit so it no longer engages smoothly. Another time I just got a new lens that had a hood that was just too tight for my liking.

Anyway, what I discovered fixed both was to put a dab of some DVD disk scratch remover compound on the hood mounting grooves. Then mount the hood and slide it back and forth several times to let the compound do its work. I'm not sure if it's the mild abrasives working or the fact that it contains some silicone and wax--maybe some of both.

Then wipe the excess off of the hood and the lens and you're done. At least in my case, the hood engaged smoothly on the lens after this treatment.

Perhaps someone knows of some similar stuff that works just as well. Eg., perhaps soft-scrub soap would work. I just happened to have this stuff so I gave it a try and it worked great.

Compare Sigma and Tamron 70-200 F/2.8

2008-11-22T11:38:00.000-08:00

Between reviews and customer testimony, I was having a hard time weighing the reported IQ advantages of the Tamron against the focus speed/quietness of the Sigma.

I originally bought a Tamron because I figured the screw drive focus on the Pentax mount would be fast enough. That Tamron was actually not a great copy so I sent it back for another (Amazon customer service was outstanding--they sent me a replacement overnight.)

Initial Quality

Here's a couple shots showing what it was I didn't like about the first Tamron. Lens #1 had visible hazing/ghosting at the focus point. Lens #2 was better--no hazing at the focus point although it does have the hazing in the OOF just behind the focus point. This tends to make portraits have a soft-light look to them. These shots are done with Pentax K10d.

(Click on the image to see the full-sized version)

Here's another example. You can see, even the 2nd (good) copy of the Tamron wasn't perfect. But I judged it good enough.

I'm making a point of showing this because I think it may be a fairly common defect in the Tamrons. So if you see this issue (mostly visible for closer subject distances), you might consider sending it back for another.

I then order a copy of the Sigma 70-200. First copy was *way* off in auto-focus--far too much even to be corrected by the K20d. Sent that back and got a second one. This one was much more accurate--AF adjustment of about -4 was required to correct it in the K20d.

So for my two-lens sample size, I got 50% for out-of-box quality for Sigma and Tamron.

Incidentally, the second Tamron I got needed -2 adjustment to focus accurately.

Image Quality and Sharpness

I put comparison shots of each lens in a gallery. It contains several chart tests so you can judge sharpness. The two are nearly identical, but I think the Sigma does better at 200mm. Focus was set manually because I didn't want to depend on perfect AF accuracy--I took a bunch of shots with miniscule focus adjustments between each shot. I did this three times with the Tamron because it's so dang sensitive, I missed the focus point doing this.

There is also what I'm calling a bokah test--an at-angle shot of a focus test chart. IMO, the Sigma easily wins the bokah test. The Tamron has a very ghosty-looking appearance in the OOF (Out-Of-Focus) region just behind the focus point. The Sigma has a much more natural-looking blurry effect. The Sigma does have some red/blue color separation in the near OOF region. However, I've yet to have that actually ruin a real-life shot whereas the weird ghosty-OOF of the Tamron does "damage" some shots.

The gallery also has a bunch of casual hand-held snaps around my yard from both lenses. All shots may not be in perfect focus, but I think it generally illustrates the very nearly identical IQ of the two lenses. Again, the red/blue color separation of the Sigma is visible in the sparkly water drop shots. The ghosty bokah of the Tamron is also evident in these shots.

All shots in this gallery are shot with the K20d which is 14.5 Mpixels. So these lenses are REALLY being stressed for sharpness.

Here's the Gallery. The image descriptions are self-explanatory:

Gallery of Images: Sigma/Tamron--70-200 F/2.8

Auto Focus Performance

Now on to focus performance. I wasn't actually expecting the focus speed to be any different. I mainly was interested in the Sigma HSM for the quietness. However, after trying it out at my son's soccer practice in low light and being pretty amazed at how well it dealt with moving subjects in such poor light, I decided to test it out.

The test setup I used at home was to set the light level to 5EV (mimics what was at the soccer practice). Put the camera on a tripod about 10 feet from a high contrast subject. Set focal length to 200mm. Set the camera focus ring to 12 feet. I also set the shutter speed to 1/100s. Used a digicam to video the camera from when I press down the shutter button to when the camera is done taking the shot. Extract the audio from the video file and analyze it using the audio package in Octave. I then was able to measure the amount of time from when the button was pressed to when the leading curtain of the shutter opened.

Here's the sound waveform for Pentax K10d + Tamron 70-200

The reason I know the leading curtain occurs where it does is those two sound spikes are exactly 1/100s separated which corresponds to the shutter speed. I ran this test a couple times and measured the average time from button push to leading curtain and it was:

Tamron focus time on K10d = 0.93s

This corresponds reasonably well with the focus times measured by Pop photo when they reviewed the K10d. Those little bursts you see between the button push sound and the camera mirror sounds are the brief sounds the lens/AF motor makes. Notice that the motor moved the lens about 4-5 times before achieving lock. You can see how each successive lens burst is shorter and shorter as the camera zeros in on lock.

The large periods of time with no sound is the camera AF system thinking about what to do next. Notice that most of the time, the lens isn't moving.

Repeated the test with the Sigma. Here's the waveform:

This time the result was:

Sigma focus time on K10d = 0.47s

Twice as fast. This is incredibly significant when you consider some of this time has nothing to do with AF. The main reason for the improvement is the Sigma lens found the correct answer with only two focus adjustments. It's not hard to imagine that the HSM motor is able to hit a given target more accurately than the mechanical screw-drive linkage for the Tamron.

This phenomonen is consistent. This result will cause me to look more closely at Sigma lenses in the future.

As an aside, I also did the test on a K20d. Results were nearly the same. K20d locked in 0.43s--virtually the same speed.

I also did the test at 2EV light levels. Tamron focused in 1.05s, Sigma focused in 0.78s. So the relative gap narrows if the light levels go low enough, but the Sigma still wins. I didn't test in bright light or with moving subjects. Didn't have time on a rainless day for that.

Aperture--is F/2.8 really F/2.8?

Some rumors running around the internet claiming the Tamron isn't really F/2.8. In reality, my comparisons with my other lenses show it really is F/2.8 (either that or every F/2.8 lens I own is wrong.) However, at smaller apertures, the Tamron is brighter than normal. Eg., F/4 is closer to F/3.5. I think this might be why people (including myself at first) suspected the Tamron might be darker than advertised. I don't consider this an issue since I only care that the lens is F/2.8 when opened all the way up.

The Sigma is spot on throughout--F/4 is exactly half as bright as F/2.8.

Handling and Construction

The Sigma focus ring is smoother and seems more tightly coupled to the glass--it's easier to make semi-predictable focus adjustments by hand than it is on the Tamron (which takes an incredibly light touch to focus).

The Sigma is effectively quick-shift. You can grab the focus ring and turn it without disengaging AF on either the lens or the camera body. The Tamron is the opposite. You have to engage manual focus on *both* the lens and the camera body. And quite often the manual focus clutch on the Tamron requires some jiggling around to snap in place.

The bayonet on the Sigma is bevelled and tapered which should (theoretically) make it easier to engage the lens to the camera throat for mounting. The Tamron has basic sharp square corners on the bayonet. In practice they seem about equal in ease of mounting. I did notice the sharp corners on the Tamron were starting to get dented, whereas this wasn't happening on the Sigma.

The tripod mount for the Tamron is held by a threaded bolt that has to be screwed tight and it's several rotations. The Sigma snaps on then tightens with a half-turn of the knob. The Tamron takes longer, but it's also feels a bit less forced.

Bart

Lens metadata fixer script

2008-01-09T22:06:00.000-08:00

There's an issue with many Sigma and Tamron lenses where the lens type EXIF metadata codes are reused by several different lenses. So when you are using EXIF-aware programs such as Bridge, Lightroom, Imatch, et al., they can't unambiguously determine which lens was used in a photo.

So I wrote a perl script that contains a customizable list of my own lenses and their codes. It steps through all of the image files in a directory and utilizes exiftool (by Phil Harvey) to read the lens code, unambiguously convert that into the name of my lens, then write that back into the lens Metadata field. At the moment, it supports JPG, DNG, and Pentax RAW (PEF). For the PEF files, it'll also modify the XMP sidecar file if it exists.

After you run the script, your Sigma and Tamron lenses should show up correctly identified.

The script assumes you don't personally own two different lenses with the same lens code--this should be very unlikely.

Anyway, here's the link to the perl script itself. You really need to know perl so you can edit it with your own lens collection:

Apple OS X version:

fixlens.zip

Microsoft Windows version:

fixlens.zip

And here's the link to get exiftool:

Exiftool

Exiftool is available for both Windows and Mac OSX. Windows users must install perl (if you know how to program perl, then you'll already have installed it!) Mac's come with perl installed.

Bart

Automatically add logo to an image.

2007-12-07T23:09:00.001-08:00

Here's a Photoshop script to apply, scale, and position my logo (which is just a scanned and modified copy of my printed name) automatically on all images. I also like to append the year the picture was shot and, in some cases the copyright mark. The script handles this automatically as well (well, sort of--you have to create the years, but it'll figure out which one to use.)

The script can be downloaded from here.
(Updated 2008-Jan-6)

I have a directory with the various versions of my logo--here's that directory in Bridge so you can see I have created several versions of the same logo differing only by year. Note the different filenames and how these match up with the filenames in the "logoFilePathList" described below.

It's easy to make all these logos--I just hand wrote all of the years and scanned them in. Then copied/pasted each year into a different version of the logo file. My logo files are all in photoshop format so they're easy to edit. Of course my logo is a simple handwritten thing--the script doesn't care what the logo is--it can be any graphical image.

The script first reads the exif year of the photo (if you've enabled that feature), then it picks one of the logo filename based on that. If you disable the year-reading feature, then it just loads your one logo file. After loading the logo file, the script scales and positions the logo based on the image dimensions and proportions. When it's all done, your image will have a new layer added to it with the logo.

Here's a sample of how mine looks:

I've recorded the script into an action so I can run it easily on a batch of images from within Bridge using the image processor tool. You can also use the action to merge the logo into your image and change the blend mode of the logo as well (for instance, I know some folks like to use overlay to blend a logo.) The script could also modify the layer styles--in my case, I've already tweaked the layer styles in the original logo file. You can do it either way.

To customize the script to read your files and scale and position the logo to your liking, you only need to worry about the declarations at the top of the script. Here's what they mean and how to change them to customize the script:

function logoWidth() {return 0.20 * longSide;}

Defines how the script scales the logo. In my case, I've told it to make the logo be 20% of the long side dimension of the image. You can change the 0.20 number to alter the size. In addition to longSide, you can also use shortSide, docWidth (the horizontal dimension), and docHeight (the vertical dimension). If you want to get fancy, you can put any formula in there that you want.

function xTran() {return 0.04 * shortSide;}
function yTran() {return 0.035 * shortSide;}
These two define how far from the edge of the photo the logo will be. In my case, I've said I want the logo to be 4% of the short side dimension from the left or right and 3.5% of the short side dimension from the top or bottom. If you used 0 in these formulas, the logo would be pushed right up against the corresponding edge of the photo.

Which edge this spacing applies to is controlled by yAlignment and xAlignment (described below).

var maxRelativeLogoSize = 0.40;
This also controls the logo scaling. It means regardless of how large the first formula above tries to make the logo, it won't be allowed to exceed 40% of either image dimension. This comes into play for very long, skinny images where the first formula above would make the logo too big.

var yAlignment = 1;
var xAlignment = 1;
These determine in which corner the logo will go. yAlignment = 0 puts the logo at the top. yAlignment = 1 puts it at the bottom. xAlignment = 0 puts the logo on the left. xAlignment = 1 putsit on the right. In my case (1,1), the logo goes in the lower right corner.

var useOnlyDefaultLogo = 0;
Setting this to 0 tells the script to use the exif year to pick a logo file (which is what I'm using). Setting it to 1 tells the script to ignore the year and just use the first entry in the logoFilePathList array (described below.)

var logoFilePathList = Array(
"/D/My Pictures/Logos/bart_hickman_logo_no_year.psd",
Array("1995", "/D/My Pictures/Logos/bart_hickman_logo_1995.psd"),
Array("1996", "/D/My Pictures/Logos/bart_hickman_logo_1996.psd"),
Array("1997", "/D/My Pictures/Logos/bart_hickman_logo_1997.psd"),
Array("1998", "/D/My Pictures/Logos/bart_hickman_logo_1998.psd"),
Array("1999", "/D/My Pictures/Logos/bart_hickman_logo_1999.psd"),
Array("2000", "/D/My Pictures/Logos/bart_hickman_logo_2000.psd"),
Array("2001", "/D/My Pictures/Logos/bart_hickman_logo_2001.psd"),
Array("2002", "/D/My Pictures/Logos/bart_hickman_logo_2002.psd"),
Array("2003", "/D/My Pictures/Logos/bart_hickman_logo_2003.psd"),
Array("2004", "/D/My Pictures/Logos/bart_hickman_logo_2004.psd"),
Array("2005", "/D/My Pictures/Logos/bart_hickman_logo_2005.psd"),
Array("2006", "/D/My Pictures/Logos/bart_hickman_logo_2006.psd"),
Array("2007", "/D/My Pictures/Logos/bart_hickman_logo_2007.psd"),
Array("2008", "/D/My Pictures/Logos/bart_hickman_logo_2008.psd")
);

This array lists all of the logo filenames. If you set useOnlyDefaultLogo = 1, then the first filename listed (bart_hickman_logo_no_year.psd) would get used for all images. You use this if you just want to put your logo on your image with no year specific variations.

However, if you set useOnlyDefaultLogo = 1, then the array tells the script which logo files to use for every year. I have the years 1995 through 2008. You can add as many years as you need to the list.

If the script doesn't find a year in the image EXIF, then it reverts to the first filename in the list.

Experiment with the parameters to see what they do. Let me know if it's useful or if you'd like some other function added.

I don't know how new your photoshop has to be for this to work. I wrote the script for CS3--I'm guessing it'll work for CS2, but I don't know how much farther back you can go. Photoshop Elements is out of luck as far as I know. It's on my list to see if there's a way to make it work for Elements.

Bart

First Trip to Arch Cape with new K10d

2007-04-04T01:33:00.000-07:00

Took our first trip to the coast since I've had my K10d. Used my DA 50-200 and Sigma 17-70. This is pretty much my "dream" kit, so I was in heaven, photographically-speaking. Most of these have been post-processed in photoshop for various effects. In particular, the weather and lighting were usually very diffuse and flat, so I like to snaz up the colors and contrast.

See shots from this trip below, or full Arch Cape Gallery (continually updated) is here

(click on any of the photos to see a larger version)

Some interesting driftwood at Arch Cape:
1/125s f/14.0 at 17.0mm iso100

A colorful, lone orange rock on the beach at Arch Cape. This rock was gone the next day--perhaps adorning some other beach.
1/50s f/18.0 at 21.0mm iso100

Favorite game at high tide--running from the turbulent surf. I play the roll of wave watcher. Water temperatures in the low 50's (F) help with the excitement! :-)
1/320s f/5.6 at 40.0mm iso100

Beach House
1/100s f/14.0 at 70.0mm iso320

I love morning portraits (especially with fill flash)--the sun makes a nice hair highlight. The range and sharpness of the Sigma are great for this:
1/160s f/4.5 at 70.0mm iso100

1/160s f/4.5 at 70.0mm iso100

1/125s f/10.0 at 17.0mm iso100

Sea foam is very interesting frozen in time as it rolled in over the rocks. Because I never know quite where the water is going, use AF.C. and select the focus point.
1/500s f/7.1 at 70.0mm iso100

Couple more foam shots here:

Hanging on to the edge of the continent. This is looking up a cliff at some trees on a steep slope above the cliff.
1/200s f/9.0 at 17.0mm iso100

Idyllic waterfall on the beach at Hug Point.
This is actually a composite of two 17mm shots--roughly equivalent to a ~14mm shot. Using the new photomerge feature in Photoshop CS3 using the "perspective" setting gives that wide-angle look.
1/100s f/13.0 at 17.0mm iso100

I found the best way to get focus on a flying bird like this was to use AF.C. in combination with letting the camera automatically choose the focus point (it's too hard to keep the dot on the bird otherwise.) Another note to self--remember to shut off the SR!
1/800s f/5.6 at 200.0mm iso250

The windblown spruce trees are a photographers delight.
1/320s f/8.0 at 21.0mm iso100

Castle Rock:
1/1000s f/7.1 at 200.0mm iso100

Interesting driftwood at Short Sands Beach
1/15s f/13.0 at 17.0mm iso100

Surfers at Short Sands Beach.
1/200s f/11.0 at 200.0mm iso125

Partly skeletized leaf--Oswald West state park
1/40s f/11.0 at 70.0mm iso640

Last shot as I got in the car to come home--fog in the coastal mountains:
1/200s f/5.6 at 200.0mm iso100

Bart
--
http://zumbari.zenfolio.com

A fun way to see SR in action on a Pentax K10d

2007-03-03T00:38:00.000-08:00

This is just a quick description of a cool way to see shake reduction in action in a Pentax K10d camera (test should work on any DSLR with stabilization, but it's really not necessary for cameras with lens-based stabilization because you can see it working through the viewfinder.) I think this is a useful test if, for whatever reason, you're suspicious the SR on your camera isn't working.

First I set my camera in manual mode, ISO100, and shutter time of 2 seconds. I used my 50-200mm lens at 200mm, but the test should work for any focal length. Then I go into a completely dark room with a heater light (or some other rapidly pulsing light source). I spend a few minutes and practice panning the camera so the heater light moves across most of the FOV of the camera during the 2s period. I was just standing up and panning my camera as steadily as I could.

Once I get my panning speed pretty good, I take a picture with the SR turned off. Here's how that picture looks (click on the image to see a larger version).

The heater light has traced out a path showing my own horizontal movement along with my random shakes and wobbles (which are considerable with a 200mm focal length)

Now repeat the same shot except with SR turned on.

Again, the light has traced out a path showing my horizontal panning, but the shakes and wobbles have been smoothed out. Pretty cool huh?

In this shot, you can also see on the left edge of the light path that it's brighter. This is apparently because the SR started out trying to attenuate the horizontal panning, but then gave up after the motion was too large. I found in more testing that it stops trying to prevent your panning motion starting at 1/5s and continuing to faster shutter speeds. This suggests the K10d SR is suitable not only for stationary subjects, but also for use when you are panning to follow a linearly moving subject (automobile, skiier, airplane, etc..)

Pentax K10d: Bright/Natural sharpening comparison

2007-01-27T22:04:00.000-08:00

I've seen some discussions about whether the Pentax K10d sharpening behaves differently in Natural and Bright modes. Since I like to use Bright mode myself, I decided I would perform a test to answer this question:

Is the sharpening filter in bright and natural modes different in any way other than magnitude?

This analysis shows what the sharpening controls are doing in each mode and should leave little room for subjective interpretation as to what's going on (of course there's plenty of room to interpret whether it's good or bad.)

The first analysis is how the sharpening transforms edges. I used my 18-55 lens at F/8 (so it's pretty sharp), used auto focus once to get the target in focus, then switched to manual focus to eliminate any tiny variations in focus between settings. I also set the camera to daylight white balance and turned off SR (it was mounted on a tripod.)

Here's what the photo target looks like (actually, this is a crop)--it's a portion of an ISO resolution test chart (as with all pictures in this post, you can click the photo to see a larger version.)

I'm only using this target because it's something I had handy that has a clean, sharp black/white edge. I then cropped further down to the black/white edge at the center of the chart. Here's what one of the crops looks like (I've zoomed in a bunch so you can really see the pixels):

I did this for natural and bright modes across a range of sharpness settings. Crops from all of the data are here:
All of the data used for this part of the analysis can be found here:
http://picasaweb.google.com/Bart.Hickman/K10dSharpnessSamples

Each photo has a code on it indicating the camera settings. Ie., "B 1" means bright mode at +1 sharpness.

Here's a plot showing luminosity versus pixel location for two of these edges. The first one is natural mode at +2 sharpness. The second is bright mode at 0 sharpness.

What you can see is, as several have already observed, brightness is about the same sharpness as natural mode at +2 sharpness. In addition, it has exactly the same transfer function characterized by an undershoot at the dark side of the step and a soft damped corner on the light side (which is what gives the images their soft appearance.)

Here's a plot for natural mode at 0 sharpness. Even here, you see the undershoot.

Just for comparison sake, here's what you'd get from the Adobe RAW converter (sharpness = 25).

Now this is a thing of beauty--very good illustration of one reason people shoot in RAW. Extremely sharp transition (approaching the theoretical limit), no overshoot or undershoot to speak of. In other words, pretty much what your eye sees. Of all the responses plotted, this is the easiest to work with in post-processing. You can modify it to get either of the in-camera jpeg responses if desired. But you can not modify either of the jpeg responses to achieve this response (at least not with any standard photoshop filters.)

To completely eliminate any subjectivity, I did a spectral analysis of these edges as well. For the technical folks out there, I perform the derivative of the edge response (aka step response), then perform a Fourier transform. Without getting too deep into theory, this gives you an equivalent spatial frequency response that would generate such an edge. I actually took several edges and averaged their fourier transforms so as to eliminate noise and artifacts from the equation.

Here are the plots for natural +2 and bright 0 in the frequency domain:

Again, virtually identical. Peaking and rolloff look virtually the same (judging from the slight extra peaking near 500Hz, natural at +2 sharpening is slightly sharper than bright at 0 sharpening.)

The horizontal axis is in Hz normalized to the image width. I think a good rough rule of thumb is you can double the frequency to get LPH. So the 1000Hz point on the graph shows the response to 2000LPH details. Nyquist is at 1936Hz which corresponds to 3872LPH. When the curve gets down to about 0.2 on the vertical axis, the details have pretty much just turned a solid color.

Do not confuse this data with actual resolution. There are too many weird non-linear and dynamic transformations being applied to most image data for this to be a reliable measure of resolution (unfortunately, I've read some camera reviews that use this method to quote resolution. It's probably accurate for RAW files from SLR cameras, but it falls completely apart for heavily processed jpegs--especially from most P&S cameras)

For reference, here's a similar plot depicting (as labeled), bright and natural modes at 0 sharpness, RAW using ACR at 25 sharpness, RAW using ACR at 40 sharpness, and something I call ideal. Ideal is simply a slightly angled edge I drew in photoshop with the "anti-alias" check box marked. I figured this is a reasonable approximation to perfection. It's interesting that RAW has a curve shape similar to ideal except the contrast ratio is muted by roughly 20% of full scale.

So that takes care of edge sharpening--natural and bright are identical in characteristic and only different in amount.

Now on to texture. I decided to just use the camera's own noise as the texture. I shot another series of photos of a smooth color over all sharpness settings of natural and bright. I did this for ISO100 (very smooth texture due to noise) and ISO1600 (coarser texture due to noise.)

What we want to do is see if sharpening amplifies this texture the same way for both modes. Here's a plot showing log10 of rms noise versus sharpeness setting. This is ISO100.

What we see is the two lines are roughly parallel. On a log plot, this means the ratio of noises between the two modes is remaining constant. Ie., both natural and bright accentuate textures identically for fine low contrast texture (ISO100 noise). It also means the noise reduction is the same.

We also see further confirmation of the equivalency of natural and bright sharpening: the noise in natural mode at +2 sharpness is about the same (slightly more) as the noise in bright mode at 0 sharpness. This confirms that the sharpness control acts equally on both textures and edges for both modes.

Here's the same plot, but for ISO1600.

In this case the two lines are parallel only up to sharpness +2 (on natural) and then they get closer together which suggests something is causing the texture in bright mode to not be accentuated at the same rate once the noise is larger than a certain amount. Perhaps it's due to NR kicking in stronger at ISO1600 when the sharpness starts to get extreme. In retrospect, I should have tried shooting a photo of an actual texture at ISO100 to see what happens. I'm pretty sure it'll duplicate the results from the first ISO100 photo and show that sharpening boosts texture regardless of natural or bright mode.

Another observation is the slope of the curves is steeper for ISO1600. I'm guessing this would be a function of the NR. I should have shot with NR off, but these results are clear-cut enough I don't want to bother.

So three conclusions:

1. From the standpoint of sharpness, bright mode = natural mode +2 (or maybe +1.8).
2. The edge sharpening characteristics are identical.
3. The texture sharpening characteristics are identical.

For me, this is good news. I never liked how other cameras seem to smooth out low contrast textures--somewhat equivalent to surface blur in photoshop. I also like that the K10d isn't mucking around with textures at higher ISO as can be seen in competing cameras. I would like to see a more RAW-like jpeg step response, but I also understand such a thing might be taxing in current-generation processing technology.

If you turn down saturation one click and sharpening two clicks, bright mode can be considered simply as a way of boosting the midtones a bit without sacrificing dynamic range.

I don't see anything that suggests natural mode should be easier to post-process than bright mode. I can unsharpen a bright image just as easily as I can sharpen a natural image. Perhaps if you turn down natural mode sharpening even further, the undershoot will completely disappear, but then you will have destroyed some texture information with such a blurry result. It would be easier to manage in photoshop if the sharpening was symmetrical (engineers like to call it linear phase) and smaller radius.

Bart

Astrophotography with Fuji F30 and Orion Starblast Telescope

2006-06-29T22:59:00.000-07:00

I was surprised to find out I could actually see Jupiter and the Galilean moons through my 4" reflector. I hooked up the F30 and shot some photos (very challenging to get good focus on stars and planets with my relatively cheezy setup). It's the same setup as in this post.

Here is a 100% crop from a shot of Jupiter that shows the bands. I did some Photoshopping to enhance the bands, but they are definitely visible. The lens specs are 114mm diameter (4.5 inches), 450mm focal length, F/4. You get roughly 30x magnification plus 3x more with the camera so that's roughly equivalent to a 3000mm, F/4 lens.

Here's another terrible shot that's got much stronger exposure so you can see three of the moons (I don't know what happened to the fourth one--I saw it last night.) Part of the problem here is shaking in the Telescope--it vibrates for a while after I let go--I only used the 2 second timer on the camera--next time I'll use the 10 second timer.

Photoshop Helper Actions (PF updated V2.6)

2006-06-18T11:08:00.000-07:00

I have three Photoshop actions which were tweaked to work with the Fuji F30, but they'll often work with any camera.

Updates:
2006-December-23 update: I now have three separate files for the three actions. Here they are:

Purple Fringe Remover. Works on Photoshop 7, CS, CS2, CS3, and Photoshop Elements 1, 2, and 3.

Radial Sharpen. Works on Photoshop CS2 and CS3.

Anti-Flare. Works on Photoshop CS2 and CS3.

The only functional update this time is to the purple fringe removing action which has been updated to work with Photoshop Elements 3.0 or earlier (I don't know if it'll work in newer versions of Elements.) I think it should also be a bit more robust for older versions of Photoshop (v7 is the oldest I know it works on.)

Anybody who tests these successfully on something else, let me know.

Photoshop installation (windows)
Photoshop users only need to copy the .atn files into their action folder then load the action into photoshop. Barts_PF_Remove_v2_6.zip also contains a file called "Barts PF Remove v2.6.psd". You can delete the this file as it's not needed for Photoshop. Restart photoshop to use the actions.

Photoshop Elements installation (windows)
Step 1:
Make sure to shutdown Photoshop Elements.
Unzip Barts_PF_Remove_v2_6.zip. There are two files called:
Barts PF Remove v2.6.atn
Barts PF Remove v2.6.psd

Step 2:
Go to path
c:\Program Files\Adobe\Photoshop Elements 3.0\Previews\Effects
and create a folder called "Barts PF Remove". Copy the two files from step 1 into this folder.

Step 3:
Go to path
c:\Program Files\Adobe\Photoshop Elements 3.0\Previews
and delete the folder called "Cache".

Step 4:
Restart elements. Because you deleted the Cache folder, give it a minute to populate the effects palette. When it's all done, you should be able to see the icon for the PF remover in the palette. It looks like this:

All you have to do now is load an image and double click this icon and cross your fingers. FYI, I have adapted the procedure found here:
http://www.hiddenelements.com/actions_whitepaper.htm

Consequently I don't know if it'll work on Photoshop Elements 4.0 or newer.

Just an FYI about opertion--if you run the purple fringe removing action on an image that has very little PF, you'll get a warning dialog saying something like "warning, no pixels more than 50% selection". Just click the OK button and the action will complete properly. If you run the action on an image with NO purple fringing at all (eg., run it on an image that's all one color) you'll get some more warnings about no pixels being selected and some commands not being available. In this case, you might as well "stop" the action because this means it's finding no purple fringing at all.

Photoshop Elements users can try this link--there are some free addons for Elements to allow it to run actions (I have not personally tried it.)

http://www.hiddenelements.com/index.htm

Purple Fringing:
download now
The first one removes purple fringing and is called "PF Remove". It's based on a radial model of purple fringing--ie., the purple fringing radiates away from the center of the image getting worse as you approach the edges (this is how purple fringing behaves in all three digital cameras I've owned.) It's also based on contrast changes of course, so there are situations where it will accidentally remove purple where it shouldn't. It's meant primarly for those situations where you have darkish foreground objects set against a bright sky, but it'll work for a variety of situations (see examples below). When you run it, keep in mind you can erase its effects in regions of the photo you know won't have any purple fringing. I fixed the problem in V1.0 that devoured color around clouds. The only errors I've seen are that it will miss some subtle forms of PF (which are hard to see anyway) and it'll attack the edges of purple flowers (kind of hard to avoid).

One other thing--you must run it only on un-cropped photos. This is because the radial modelling of the lens assumes the entire original photo is present. Photos can be scaled in size, but not cropped.

Here's a typical before/after situation where you have some purple fringing on some foliage and sky in the background (click on the image to see the 100% crop). The crop is done after running PF remove of course.

So you can see the action has removed the purple. You might also be able to see another cool feature on this action and that is it put the green back where the purple was. It's hard to tell because the foliage is low saturation, but you'll see the difference if you turn off the "PF Replace Color" layer (see below).

When you run the action, it creates two new layers. No changes are made to existing layers! So you can always undo the effect by delete these two new layers. An example of these layers is shown here:

The first layer is a Hue/Saturation adjustment layer labelled "PF Removed". The mask on this layer designates the areas of the image that have dark-to-light contrast changes radiating from the center of the image plus some allowance for flare (which is omni-directional). Any purple in the white parts of the mask gets removed (desaturated.) The purple fringing on the Fuji F30 is really PURPLE. This Hue/Saturation layer is targeting that color. If you have a different model of camera with a different color of purple (ie., many camera's have PF that is more blue), then you can fiddle with this adjustment layer (or, better yet, fiddle with the appropriate step in the action) to customize it for your camera.

The second layer is an image layer called "PF Replace Color". This layer will usually look mostly transparent, but if you look at it closely, you'll see it has some non-transparent information on it. The blend mode of this layer is "color" and it is the layer that's replacing the color that belongs where the purple used to be. You can manually paint on this layer to do any further color fixes (for instance, any purple that wasn't removed or any colors that weren't properly replaced.) This layer also shows you exactly where purple was removed so you can make sure no desired purple was removed. You can turn off this layer to see how it would look without the color replacement.

Radial Sharpen--improves lens edge softness
download now
Most P&S cameras have this at some focal lengths to one degree or another. My Z750 has it pretty noticeably in full zoom. The F30 has a slight amount at full wide angle. If you want to get rid of it, I have an action that'll do it--it's called "Lens Sharpen". This action sharpens in the radial direction progressively more as you move away from the center of the image--so you need to run it on un-cropped images (resized is okay.) I tuned it primarly for 8mm focal length (full wide) on the F30 which is far and away the most common zoom setting I use. In the case of the F30, I think the edge softness of that camera is mostly in the radial direction. I think this is at least partly because Fuji corrects barrel distortion in the camera which tends to stretch, and therefore soften the edges in the radial direction.

When this action runs, it creates a special type of unsharp mask layer that is sharpening in only the radial direction with progressively larger radius as you approach the edges. So it has no effect near the center of your image. By default, the opacity is at 30% because that seems to look good. (Note that the blend mode is linear light.) You can change the opacity to increase or decrease the edge sharpness to taste. The sharpening layer is called "Radial Sharpening".

Mega-blow out flare
download now
There's a third problem I come across more rarely than purple fringing that occurs when the dynamic range of the scene is huge--ie very dark foreground and very bright background. I'll call it lens flare, but it can also be a function of the CCD. It can be minimized by turning down the in-camera contrast, but the F30 unfortunately has no way to turn down contrast and the F30 seems more extreme than other cameras--at least more so than my Casio Z750. I've encountered the problem infrequently enough that I haven't had a chance to test this action very much. The action I use to help remove this problem is called "Anti-Flare".

Here's a crop from a photo that has this problem in a small area:

The flare from the bright sky is obscuring the leaves. When you run the script, it creates two new layers similarly to the PF remove script:

The first layer is called "Anti-Flare Curve". This is controlling the strength of the anti-flare effect. Photoshop users can adjust the curve itself to fine-tune the effect for the specific situation which can be quite variable. I don't think Photoshop Elements will allow you to actually adjust the curve, so you may have to just adjust the opacity of that layer. Here's the result for our example.

Not perfect, but good enough. The second layer created by the action is called "Anti-Flare Replace Color". I just sampled some nearby leaf colors and painted sloppily in the flare location to put the green back in:

Telescope/Astro Photography w/ Point & Shoot Camera

2006-06-06T00:33:00.000-07:00

I took some moon photos tonight with the Z750. The weather was perfect!

This first shot is how it looks just with the Z750 at full zoom:

Here's the moon shot through my telescope which is an Orion Starblast reflector telescope, 4.5' diameter, 26x zoom which I guess means it has an equivalent 35mm focal length of about 1300mm. It's probably the cheapest reflector you can get. (click on the image to see larger size.)

You can see some softness in the moon horizon--this seems to be a limitation of the combined Z750/telescope optics I haven't been able to overcome.

However, Picture Cooler + some Photoshopping does a pretty good job of correcting it. Here's the PP'd image. (click on the image to see a larger size).

Here are some photos of the telescope rig: