How do I take/edit great photographs?
July 9, 2005 6:30 AM   Subscribe

Okay, so everyone loves DaShiv's photographs of SF Meetups, and for good reason. They are incredible. I can take decent photos, but they're obviously amateurish. Here's my question: How do I take it to the next level? How can I get the colors to look as brilliant as DaShiv's, for instance (just an example; I'm sure that there are many other styles of photography that I am not aware of)? What equipment do I need? What software do I need? What type of training is required? BTW, I run a Mac and have a little cash to burn. Thanks in advance.
posted by SeizeTheDay to Sports, Hobbies, & Recreation (26 answers total) 5 users marked this as a favorite
you can get at least part of the way there using gimp (for free). look at the colour balance and levels tools. colour balance will remove colour casts from odd lighting conditions and the level tool will correct your exposure settings.
posted by andrew cooke at 6:53 AM on July 9, 2005

DaShiv summarizes what he does in this post.
posted by driveler at 7:03 AM on July 9, 2005

Good lenses. The nice blurred backgrounds and the sharpness in the portraits come from using a fast lens that can have the aperture opened right up.

There are other ways, but if you want saturated colours and soft backgrounds like that, buy a fast prime lens and use slide film. Or get a digital body, and photoshop the colours. But the lens is the thing. (Also not using full-on blinding flash is important)
posted by bonaldi at 7:03 AM on July 9, 2005

buy a fast prime lens and use slide film.

Also, get religous about exposure. Slide film isn't nearly as tolerant as print film, so you need to nail the exposure.

The bonus: Digital isn't very tolerant either, so you can transfer that skill to digital, when E-6 is all but gone. Note that this is stuff you really can't correct for later -- you can alter the information present, you can even change the relations between them, but you can't easily add information that's not there as a result of under/over exposure.
posted by eriko at 7:15 AM on July 9, 2005

This looks like a rangefinder camera, and this appears to be a Canon, but I don't know Canons. Regardless, he is using very high quality lenses (there is no substitute for this), and because they are portraits, he's blowing out the backgrounds by setting a very short depth of field, which can be achieved by opening the aperture and using a faster shutter speed. Your focus has to be right on, and his usually is. Blowing out the background makes the subject jump out and allows you to focus on hos brilliant the colors are. This depth-of-field control is one area where most digital cameras fall far, far short of film cameras.
posted by planetkyoto at 7:16 AM on July 9, 2005

Macgimp - free

Photshop elements
- around $100 as I recall

A couple of photoshop tutorials
Four steps to better photos
photoshop roadmap
Hue/sat tutorial

But quite a bit is the original shot. Candids, well shot, lots of shots.

Good popular photo commentary

This is a pretty good grasp of composition

That'll get you started.
posted by filmgeek at 7:36 AM on July 9, 2005

You say you want to take it to the next level, but you don't say what level you're at. And maybe the above advice is exactly what you're looking for, but I have the same question, and all that technical and equipment detail just makes my eyes glaze over.

I've really improved my photography, though, over the past 6 months. My "technique" has been to average about 1000 photographs (taken) per month. It's not huge, but it's enough to keep it on my mind, motivate me to explore the controls of my camera, and figure out what feels like progress. It was even enough to make me pinpoint what I wished I had in the camera I was using, and get me able to identify what I wanted in my next camera.

I'm sure there are a ton of fancy tricks for optimizing color balance and mastering filters, and getting the very best "material" for post-processing by deeply understanding all of a manual camera's controls. But you can get beautiful, brilliant, interesting, original photos out of a point-and-shoot if you make your first priorities composition and a better awareness of light.
posted by caitlinb at 7:40 AM on July 9, 2005

The tools vs. technique thing is always contentious.

Obviously you are not going to be able to take good photos with a poor lens or a camera with broken exposure metering but modern cameras are, for the most part, all adequate in terms of technology. However, you could be using a medium-format camera with a digital back and the fastest lens available (figure around 20k for the package) and still wind up with garbage. Composition and being able to "see" the results are much more important than gear.

This "seeing" only comes with practice and the single biggest thing you can do is take pictures -- take *lots* of pictures -- but not willy-nilly. Pay attention to what your doing and compare the results, while looking at the histogram, shutter speed, ISO, aperture and any other information your camera makes available, to what you "saw". Then, if possible, do it over until you get what you get the desired result. Eventually you will get a feel for the best settings in any given situation. This is why digital is such a great learning tool, you get the instant preview -- pros used to use Polaroid for the same reason.

Ignore software as much as possible. Photoshop is wonderful but if the data isn't there in the first place it's worthless. You can only manipulate what you have to start with and most serious photographers regard the best picture as the one requiring the least post-processing. Cropping and a gazillion pixels are not your friend, while they can save a poorly framed shot, if you rely on them you will never learn composition. The only exception to this is b&w, good software will convert your images to b&w much better than the camera will.

Frequently overlooked is monitor color calibration. You cannot evaluate your photos if the color you are seeing isn't accurate. This is why you might take a picture, adjust the levels in Photoshop and then wonder why it looks like crap printed or on another computer.
posted by cedar at 8:16 AM on July 9, 2005

The first thing I'd ask is: what sort of camera are you using now? If you're using a digital or film point and shoot with a really long (and probably really slow) integrated zoom lens then you're almost certainly required to use a flash on every single shot, which is over exposing your subject and blacking out the background.

Want to take pictures that will make you want to quit your job? Go buy a Pentax K-1000 or ZX-M (these are film slrs). It will cost you about $100. Buy a Pentax 50mm 1.4 lens to go on the body. This will cost you $50-100. Buy a roll of slide film. If you really want to be amazed, make sure it's Velvia (made by Fuji).

Make sure and get mailers for developing the film, which will almost certainly be your most economical option. The slides that you get back will be many many times better looking than anything that a cheap point and shoot is going to create. Make sure you look at them on a light table with a loupe. You'll want smelling salts nearby.

If you're using digital then get or borrow something like a Digital Rebel or D70. Use a prime (not a zoom) lens. This outfit will cost you less than $1000. Make sure you don't use the flash. The narrow depth of field and natural light will almost instantly give you better pictures.

The trick with digital SLRs is that the physical size of the sensor often makes more of a difference than the number of pixels in the quality of the picture. Your average 8mp ultra compact camera has a sensor the size of your pinky nail. Your average digital SLR has a sensor the size of a stamp. The larger size means less noise which almost always means better quality.
posted by bshort at 8:40 AM on July 9, 2005

By the way, the pictures on my site, which can be found in my profile, are almost entirely taken on a Nikon D70. If you have questions about specific pictures or lenses or techniques let me know.
posted by bshort at 8:44 AM on July 9, 2005

Your average 8mp ultra compact camera has a sensor the size of your pinky nail. Your average digital SLR has a sensor the size of a stamp. The larger size means less noise which almost always means better quality.

This sounds quite reasonable, but I would love to see some technical details about it - from the electrical engineering perspective, rather than the photographic.
posted by Chuckles at 8:54 AM on July 9, 2005

Here are a few things I found on the issue
posted by bshort at 8:59 AM on July 9, 2005

Chuckles, C/Net has a brief overview, Creative Pro goes into a little more detail and Digital Outback goes into even more detail with a handy comparison chart.
posted by cedar at 9:05 AM on July 9, 2005

Larger sensors are the secret to why 5 megapixels from a digital SLR beat 5 megapixels from a consumer digicam. To spread the same number of pixels over a larger sensor area, the pixels (technically, photosites containing diodes) must be bigger. These bigger photosites gather more light, so they produce less-noisy images, capture greater dynamic range, and perform much better at high ISO settings.

that seems plain wrong to me (it's on the cnet site; i'm an ex-astronomer, now programmer, and i write software that processes data from ccds, so understanding noise from these things is part of my job).

the signal to noise on a ccd depends mainly on the number of photons detected. so it is influenced by (1) the number of incident photons and (2) the conversion efficiency of the ccd.

neither of those depend particularly on the physical size of the detector - you can see this by imaging a film (or overhead) projector, shining light on a screen. if you move the screen further away, the image is bigger, but dimmer. the same amount of light is spread out. now consider how that applies to ccds: assuming that you are taking the same size image, you can use either a large ccd further away, or a smaller ccd closer to the lens. in either case, assuming that they have the same number of pixels, a pixel gets the same amount of light.

in other words, the lens controls how much light comes in. if you have a "fast" lens, you get more light. this can then be focussed onto a small detector or a large one - in either case, assuming you do the optics correctly, you get the same amount of light on the detector. you can have less light over a larger area, or more light over a smaller area - it adds up the same.

having said all that, there may be technical details that influence this. for example, it may be that each pixel has a "border" that is a fixed size. if you make chips larger, the fractional area covered by the border (for the same number of pixels) is less, so you have a more efficient chip. it may also be the case that a larger detector, which implies a longer focal length for the same final picture, somehow allows the lens to be faster. but neither of those details match what cnet claims.

i'm wary of saying that this is completely wrong, because sometimes small details are important, and sometimes i just stupid (photo cameras are not identical to cameras on telescopes). but, even so, i would guess that large detectors are used more because they allow other parts of the camer to work better, rather than to "catch more light in bigger pixels".

for some support to what i'm saying see this technical article on ccds. if you search for "pixel size" you end up down near the bottom of the page, and the discussion is all about selecting the right pixels size to match the optics. there's nothing about efficiency. at one point they say A 12-mm pixel used with a 100-mm focal length achieves the same spatial resolution as a 24-mm pixel and a 200-mm focal length, but at a lower cost, since bigger pixels usually implies higher prices. which is the idea i was illustrating with my projector example.

and, reading cedar's last link this is what they say there too - detector size is chosen to match the optics. there is nothing about larger detectors being more efficient "per se", simply that expensive lenses tend to be made for a certain focal length, so you need a detector the same size as a 35mm film frame.
posted by andrew cooke at 9:30 AM on July 9, 2005

sorry, "made for a certain detector size, if you want the same focal length, so you need..."
posted by andrew cooke at 9:32 AM on July 9, 2005

another way of saying all that is to rewrite the cnet quote to read:

Larger sensors are the secret to why 5 megapixels from a digital SLR beat 5 megapixels from a consumer digicam. To use the more efficient lenses developed for 35mm SLRs at the same focal length (remember, f number is relative to focal length) you need a larger detector. These lenses gather more light, so they produce less-noisy images, capture greater dynamic range, and perform much better at high ISO settings.
posted by andrew cooke at 9:37 AM on July 9, 2005

DaShiv: most of the photos were taken with a Canon 20D and the EF 85mm f/1.2L, handheld at ISO 800 wide open. The camera body is a fairly unremarkable workhorse, but that lens is magical.

Lens: about $1,500
Camera Body: about $1,500

Naturally it's not all about money. $3000 alone won't make you a great photographer (and I'm sure you could do a lot with less than $3K). But it is kinda uselss, after a certain point, to strain to try to get more out of that $500 "prosumer" digital camera you thought would take you to the next level. You've got to start thinking on a different level if you're going to be a badass like DaShiv. You simply cannot get a lens as fast as this for much less - and the speed of the lens means a lot.

Lower f-number = higher lens speed.
High lens speed = lets more light in.
More light in = higher shutter speeds.
Higher shutter speeds = sharper shots.
More light in also = no need to use flash.
High lens speed also = larger aperature.
Larger aperature = shallow depth of field.
Shallow depth of field = those cool shots where the main subject is the only thing in sharp focus in the entire frame. Very dramatic. A key differentiator between "pro" looking shots and "amateur" looking shots.

If you hang around DaShiv you'll also notice that he almost never takes a single photo at a time. He shoots off 3-8 in automatic-rapid-fire the way you or I would snap one. This gives him a better chance of catching the human face "just-so," nailing that momentary facial expression exactly right for a nice portrait. I don't know actually if he's into auto-bracketing, but you can set most professional SLRs to automatically shoot 3-5 photos across a range of exposures every time you press the button. This, again, gives you a better chance of getting the exposure "just-so." A big chunk of the process is throwing out the shots that don't work so well. He also works tirelessly, taking hundreds and hundreds of shots at the average meetup. He's intent. He's not shy. He's learned how to be subtle. And he uses a strong enough telephoto to be able to hang back away from the action and pick off ripe moments at will.

It's much harder to take the 100s-of-pictures shotgun approach with a consumer level camera that requires 1.5 seconds to focus, warm up the memory buffer, and take a single photo. And forget about film.
posted by scarabic at 10:08 AM on July 9, 2005

Andrew, if I take my NikonCoolpix 990, set it to TIFF mode, choose a manual exposure of, say, 1s, cover the lens completely before releasing the shutter and then look at the resulting image file what I see is not uniformly black.

My understanding is that this is due to small variations and imperfections in the sensitivity of the photosites and slight variations in the integrated circuit traces and devices used to read out the data.

These variations and imperfections have more impact on smaller and higher resolution sensors because they have smaller photosites for the individual pixels, and an imperfection is going to have a more significant impact.

Let's say the average size of an imperfection is 1 square and the average density is one imperfection in every 100 squares.

If the size of a photosite for an individual pixel is 100 squares, then most photosites will have imperfections, those imperfections will amount amounting to 1% of their area, and lets assume that amounts to a 1% error for each pixel.

If the size of a photosite for an individual pixel is 10 squares, then only 10% will have imperfections, but those with imperfections will have a 10% error margin.

So, larger sensors are good both because they can collect more light resulting in a better signal, and because they are more forgiving of imperfections
posted by Good Brain at 2:29 PM on July 9, 2005

that's interesting, but again it doesn't fit with that strange explanation that cnet gave (it's not that they're capturing more light, but that you're hoping that the pixel scale is much larger than the scale of impurities, so that increasing the scale gets you a more "typical" collection of impurities in each pixel, right?).

for optimum image quality we do the following:
- take many pictures with no light (sufficient that the average has negligable scatter from readout noise in the amplifier)
- average those together. this gives you a "bias frame", which is the zero level for your images. typically it has structure due to readout electronics.
- do the same with an exposure of something evenly illuminated and "white". this is either a big white screen on the inside of the telescope done (not very good) or the dusk sky (typically astronomical images are taken through fairly narrow band filters, so "white" isn't so important)
- subtract the bias frame from those and average them together.
- that gives you something that should be flat (or at least smooth, since it will die off at the edges from vignetting). fit a smooth surface to that (or 2d median filter it) and divide the original by the fitted surface. that gives you a "divisive flat field"; something that is on average 1, but has lower and higher values that reflect pixel to pixel variations in sensitivity (what you are hoping to reduce with larger pixels).
- take images (at last! - the above process happens at the start of every night).
- for each image, subtract the bias and divide by the flat field

that removes the need to worry about pixel to pixel variations, but i suspect that it is (1) too much work for photographers and (2) unlikely to work well because you probably use your camera in very different conditions (detector temperature, for example) between calibrating and shooting photos. but it might be a useful approach for pushing to very high iso numbers, for example.

for camera ccds you're probably getting close to saturating, so you'd also need to correct for non-linear response (as the ccd saturates it gets less efficient) - this is necessary in astronomy in the infra-red (where the sky glows brighter than the stars). you have to do extra exposures (of different times) to work out what the response function is and correct before calculating the flat field. again, this is probably not necessary if you're using an approach like the above to push iso number with low light levels.

if you or anyone else tries doing the above, i'd be curious to know how it works out. there's a freely available software package, iraf, that will do these steps for you, but the learning curve is pretty steep (and they are about to cut support for users...).
posted by andrew cooke at 2:51 PM on July 9, 2005

"telescope done" should be"telescope dome", sorry (because "done" is a word that can really confuse).
posted by andrew cooke at 2:55 PM on July 9, 2005

A longish lens with good sharp optics which is opened up a bit makes those beautiful sharp portraits with nice blurry, unobtrusive backgrounds. Nice work Dashiv.
posted by caddis at 7:50 PM on July 9, 2005

There has been some awesome advice here, folks. Thanks. I guess my goal now is to find a lightly used DSLR (I like the Rebel), a decent lens, and a tutorial for Photoshop.
posted by SeizeTheDay at 8:13 PM on July 9, 2005

Yep, what you want is a large-aperture (fast) prime on a DSLR. Here are some I took at a Seattle blog meetup with a Canon EOS 20D and a Canon 50mm f/2.5 macro lens: 1, 2, 3. No flash was used for any of these photos, which were taken at ISO 1600.

You don't need a $1500 camera and lens; a Canon Digital Rebel ($700) and 50mm f/1.8 prime ($75) represent a good place to start.
posted by kindall at 11:21 PM on July 9, 2005

Thanks to bshort, cedar, andrew cooke and Good Brain for helping me out with my hijack.
posted by Chuckles at 7:53 AM on July 10, 2005

Wow -- didn't find out about this AskMe thread until someone commented about it on MeTa. A few responses:

"Vibrant" colors: I'm actually a bit puzzled by this, since Canon DSLR's (reputedly) have more muted colors than their counterparts from other brands. I don't bump up the saturation levels during RAW conversion, either. About the only thing I can add is that indoor lighting is often surprisingly colorful and interesting when you can manage to capture it without washing it all out with flash. Other than exposure issues (shoot at high ISO's and don't be afraid of noise), the other main concern is always the color balance to render the skin tones naturally when it's awash with all that funky light.

Fast lenses: Invaluable for portraits, and kindall's right, a 50/1.8 is a great and cheap way to get started with people (that's the way I started too). And his samples show that you don't need an f/1.2 lens -- an f/2.5 lens in the right situation can also blow out the backgrounds pleasingly. On the other hand, zooms are almost always in the f/4-f/5.6 range, and usually need to be stopped down more for sharpness more than primes do. All photographers should have a fast prime if they want to take portraits, whether it's a $75 lens or a $1500 lens.

The numbers: scarabic's also right in mentioning that I shoot a lot and only post a handful. For example, I probably took over two dozen shots of striatic but didn't post a single one because I didn't like how any of them turned out: the red light over the booth was pointed right at him to put odd colors on his face, his hat made lighting his eyes difficult, I didn't catch the right expression, etc. If you shoot lots and are picky about what you post, people will think that you're a better photographer than you actually are. It works for me, anyway -- I'm still learning, just like you.
posted by DaShiv at 12:04 PM on July 10, 2005

It's not the lens, and it's not Photoshop. The number one mistake photograhers make is not being close enough to the subject. Stand close. You can't stand close enough. And while you're at it, make sure the subject's eyes are in focus. If humans can't see the eyes, they will not trust the photo.

Photoshop is for fixing bad photos.
posted by jranft at 8:32 PM on July 11, 2005

« Older Where to stay in Maui   |   lap-tøp no-døz Newer »
This thread is closed to new comments.