Flutterby™! : Photo tips 2002-09-17 16:26:33+00

Photo tips

2002-09-17 16:26:33+00 by Dan Lyke 25 comments

In the Canon 1Ds thread, Meuon mentioned trying to learn more about good photography. I'll start:

If your digital camera has a "histogram" feature, make friends with it. Ansel Adams' "Zone System" is all about manipulating contrast curves to capture the shadow detail and the highlight detail, and when the scene was too high contrast, to consciously decide what throw away.
If it doesn't have a histogram feature, make friends with the closest thing you've got to a spot meter. Learn how far under and how far over your camera catches detail.
Know this, then learn to trust the automatic metering. Don't always be an exposure geek.
Learn about The Golden Mean. If you can't assimilate all that, approximate with "2/5^ths", and try to put points of detail on those places in the photo.
Some people think that the eye scans into frame on the upper left side, goes across to the right, down, back left, up, and then exits the frame just below that point. Try to set up the lines and dominant shapes in your images to encourage that flow. When that gets boring, try to make the eye scan a different path. Remember The Golden Mean as you do this.
Backlight. This counteracts everything you've ever been told about taking pictures with people in them, but once you understand exposure, make sure that people are always standing with their backs to the sun. Give them halos. See the light through their hair. When necessary, use fill flash, even in the middle of the day.
Light from the side or the bottom. Yes, I know, I said "backlight", but people like images that show them things they don't normally see. They always see light overhead.

Anyone else?

I wholeheartedly agree with Dan about the compositional importance of the "The Golden Mean." Here is another description of it, and some examples of my own compositions which I think tend to follow it. Note that it can be applied to either horizontal or vertical compositions:

As for the available histograms in digital cameras, I'd also like to see improvements to the user interface to them. Realistically, the interface should be able to provide real-time "what you see is what you actually get." In traditional photography, those kinds of predictive exposure results could only be obtained with lots of experience with the zone system, your specific metering system, the chosen film, the final presentation medium, and the transfer functions between each.

The sidelighting thing is highly, highly underrated. The slanting rays of the evening sun are extremely dramatic (D. W. Griffith referred to this time of day as "the golden hour" because of the incredible quality of light) and I've found that when doing artifical lighting, simulating this kind of lighting can create great results (see this photo for instance -- I took that with a 25-watt bulb in a desk lamp). Other examples of my work here.

Some other tips: Bracket exposures. Get Photoshop and learn how to use it, I mean really learn how to use it. (Among other things, using Photoshop it's possible to take a group of bracketed exposures of the same scene and combine them so you get plenty of detail in both shadows and highlights.)

Some excellent books for budding photographers include "Learning to See Creatively" and "Understanding Exposure" (both by Bryan Peterson) and "Light: Science and Magic" by Fil Hunter and Paul Fuqua. The latter's pricey, but worth it.

For many of my landscapes, I also prefer sidelighting as it brings out many additional details. In addition to the angle of light with natural sidelighting, I also find that I like the warmer ambient color temperatures during those periods of the day (early morning, late afternoon) which can produce the desired compositional sidelighting and increase the perceived color saturation.

Notwithstanding spot metering, I generally underexpose 1/3 or 1/2 f-stop to increase the color saturation in normal scenes (my own personal preference). Depending on the foreground subjects, I may underexpose 1-3 f-stops for maximum color saturation during sunsets or extreme backlit conditions where no fill flash is desired. I also make extensive use of polarizing filters to reduced reflected light and to also increase the resulting color saturation.

To Jerry's point, I would like to become better with Photoshop. If "making photographs" really is art, then adding another tool to the palette is simply an extension of this. I doesn't matter where the manipulation occurs; through the viewfinder, through the exposure, through the image capture (digital), through development, through scanning, through printing, or via post-processing in Photoshop. Art is art ...

Ansel Adams manipulated the hell out of his photographs in the darkroom. Photoshop is merely an extension of this, and it doesn't smell of chemicals either.

RE: Bracketing. What I am about to suggest may not be possible, or it may already be part of some prosumer or professional digital cameras. What I'd like to see is selectable bracketing where the actual "exposure" is made only one time but, for sake of example (3) separate images are stored into three separate memory buffers based on modified (bracketed) output to each from the CCD sensor. That way, you'd get perfect pixel-for-pixel registration of the resulting images, which might make any subsequent masking with a tool like Photoshop easier?

One caveat, it would seem that using this approach you wouldn't be able to take advantage of "turning up" or "turning down" the sensitivity of the pixels in the image sensor (i.e. changing the "exposure"), thereby preventing the capture of additional image dynamic range.

It's a good idea, but it wouldn't really do much as far as bracketing goes. All you're really doing at that point is scaling the numbers you got from the CCD. But if you've blown out the highlights in one part of the CCD, you can't get them back by multiplying them all by .9 because they have already been lost, and if some parts of the image are too dark to register at all, you can't bring them out by multiplying them all by 1.5 because they weren't captured to begin with.

The sole exception to this, I suppose, might be if the CCD had some "headroom" beyond the values the camera usually recorded in files. That is, a zero in the file might mean that the voltage in the CCD was below some non-zero value, and a 255 might mean the voltage was above a certain value. But camera manufacturers wouldn't do this, because they want to get the most dynamic range and the best signal-to-noise ratio they can out of their sensors. The only really proper way to do bracketing, then, is to vary the amount of light falling on the CCD.

I'm not completely up on CCD technology, but my impression is that they generally have a lot wider range than the A/D converters that read off them are using. The Canon

cameras get at least 10 bits per pixel, others get 12, and if someone could figure out ways to move all of that data around cheaper I'm sure it'd be possible to go higher. But moving that data is an issue, when we're getting 100k/sec on the CompactFlash

cards, and when image processing takes so much compute (read: battery).

One possible solution would be to deliver two images, a contrast masked image, and the mask. You could derive the full information from the image by adding the two.

Which brings me to: I haven't read the contrast mask articles above, but unsharp masking is basically taking a blurred version of the original and subtracting it from the original. Depending on your blur, this ups local contrast while reducing global contrast. Time to see if I can get Film Gimp working, since it has the resolution to actually do interesting stuff.

Yeah, some CCDs support 10 or even 12 bits, but what you do (assuming you're converting to a format that only supports 8 bits per channel, rather than going directly into Photoshop, which supports 16 bits per channel) is just chop off the lowest 2 or 4 bits, since that's where all the noise is anyway. The reason they don't do more than 12 is that additional bits would be meaningless; CCDs just don't have the required level of accuracy.

The problem I have here is that I suffer a jargon and general knowledge deficiency. I know what bracketing exposures is on a normal camera, but I don't even know how one would go about doing it on my (reasonably feature-laden Olympus) digital. I'll tell you this: If it involves manipulating the internal menus before taking each shot, forget it. My subject would be dead by the time I got those right; they are cryptic and the buttons are small and my fingers are large.

I don't know how to read or use a histogram. I presume it is some measure of the relative quantities of "light" and "dark" in the image - at least that's what the curve I sometimes stumble across in Painter or Photoshop seems to be saying - but I am never sure what tinkering with it will do.

I wouldn't mind spot exposure checks. Even the little needle in a Pentax body would be fine. After a while using that, I realized it was doing spot checks on the light coming into that little focus circle in the middle of the view, which meant I could wave the camera around a bit before taking the shot and get a general idea of what was going to happen.

But none of this addresses my main problem with the Olympus: Bad viewfinder, very bad. I can't line up pictures well using the little power-drinking screen (I have shaky hands and they're much steadier when the camera is actually up against my face), so having a real look-through viewfinder was one of my only real purchase requirements. Unfortunately its parallax is ENORMOUS (that is, it doesn't see what the lens actually sees, but has a huge offset) and it doesn't even realistically show the correct image area! This results in me having to discard two-thirds of my shots because I didn't get what I thought I would get in frame.

When are they going to come out with a digital camera that exactly mimics the behavior of a real camera - i.e. change your "f-stop" equivalent by turning a dial around the lens, zoom ditto? Is there one now for under a thousand dollars? Because if I ever saw one my Olympus would be headed straight for eBay.

Add to that mimicking comment: And it would need to be true single-lens-reflex. If they can flip up the little mirror in front of film, then they can flip it up in front of a CCD too.

The Canon 1Ds, D30, D60, the Nikon D-1, and there's a Fuji with a Nikon mount too (the S1?), are all traditional SLRs with interchangeable lenses that just happen to have digital backs. But all of those are at least $2k+ range, all take external lenses at additional cost. And then there are the ones which take medium format, which start at about $14k...

You might look at the Olympus E10 and E20. I believe both of them are optical through-the-lens SLRs which are lower priced than the aforementioned (a quick search puts the E10 at < $1200 from reputable dealers, with lens), and there are a few cameras nowadays which have small lower power consumption viewfinders placed where a traditional SLR eyepiece would be.

As for the function of the histogram, it indeed shows you the distribution of the light and dark on the image. If it skews high, you overexposed and lost highlight details (like cloud texture). Low, you underexposed and lost shadow detail. Lots of zero on either ends, you can stretch the image brightness range to make the low contrast more apparent. No zero on either ends, it might be time to get the tripod out and take multiple exposures at different settings, compositing them later, to get the higher contrast range (ie: bracketing).

On the application I'm working on today, our lighting challenge is to get nothing in the middle, everything either on a line up at the bright end, or in a wider cluster down at the bottom end, so that we can extract elements of the scene using luminance keying. Another challenge in the same rig is lighting so that we get a reasonable blue distribution, with everything else on a line at the bright end, for testing our chromakey setup to make sure that the blue bits are all shadowed and lit appropriately, with the bright bits blown out, before we add other colors into the scene.

To Dan's point ... If the CCD's offer greater native dynamic range than the A/D converters to which they write, then my previous post about the possibility of real-time exposure bracketing may actually be possible from an engineering perspective. The crux of our question is: how much dynamic range is actually captured by the CCD itself?

Which leads me somewhat naively to ask: how is the "shutter speed" actually executed in a digital camera? Is there actually a physical shutter, or is the CCD "turned on" for the required exposure duration? I might naively further ask whether the sensitivity of the CCD can be altered in much the same way that the amplication factor of a simple amplifier circuit is controlled by a bias voltage?

RE: Jerry--what about the "RAW" image formats? Do they retain the full 10-12 bits of raw CCD data? Are the "RAW" formats proprietary to each camera manufacturer? If not, perhaps Photoshop will soon read this format?

I too lament this current shortcoming. It would be wonderful to have a "what you see is what you get" type of viewfinder in all digital cameras. Surely the manufacturers could figure out an economical way to take the output of the CCD and display it in the viewfinder, eliminating the parallex issues? The 1.8" display at the rear of the camera can stay, for quick review purposes ...

Oh yeah, I also realized I've been using "CCD" up there as interchangeable with "sensor". The D30 and D60 use a CMOS sensor of some sort. Just so that Aaron or someone who really has a clue only has a little ammunition for deriding me for my ignorance.

Most RAW image formats do store the entire 10-12 bits of data from the CCD, assuming the sensor supports that many useful bits. The RAW formats are all proprietary, but the manufacturer provides software to export as, say, 16-bit TIFF, which can then be read by Photoshop without any problem.

Frankly, in current cameras, I'm more concerned that two thirds of the data is interpolated. Even a 2CCD camera that used the second CCD for luminance at every pixel would be a great improvement over the Bayer-pattern sensors used now.

Some current digital cameras do have electronic viewfinders. The 5MP Sony does, I believe, as does the Nikon Coolpix 5700 and the Minolta D7.

The sensitivity of the CCD can indeed be altered electronically. This makes it noisier, just like you'd probably expect. But most cameras allow you to set ISO 50, 100, 200, 400, and sometimes 800.

Well, I am sorry to come late to this table but I'd place my photographic level near Meuon's where I've had classes did a lot of Black and white photography ages ago and darkroom work because it was fun but lost interest along the way. I just recently started getting interested in photography again with digital media. I loved the pictures Meuon took at buring man and think that his model 707?? or this one might be the right one for me to get more serious with. Suggestions for the novice? Oh BTW I think I perfer taking shots of people vs landscape if that slants modle preferance much.

Just some quick comments: My Sony 707 has a very decent viewfinder, plus a screen, and can store big uncompressed TIF files (not JPG's) with much more data. How many bits? Ain't got a clue. Battery life has been good.. even with flash usage. As for adjusting F and exposure.. it's a little jog wheel and it works great in manual or semi-manual mode. I'm learning. I'm impressed with the camera.