How can I improve contrast of photos at the time they are taken without resorting to curves and levels tools in photoshop?
December 12, 2007 10:59 AM   Subscribe

PhotographyFilter: How can I improve the dynamic range of photos at the time they are taken without resorting to curves and levels tools in photoshop?

I understand how to improve the contrast, or dynamic range, of a photo in post processing using photoshop by playing with the curves or levels tools. If you think in terms of the histogram of the image, what I am talking about is how to take the photo such that the histogram fills up the range without clipping.

Specifically, suppose you take a photograph of a crumpled white piece of paper. Using the manual settings, I can completely avoid clipping, but I get little in the way of contrast across wrinkles, etc. In other words, the histogram fills a tiny portion of the total range. What I want to do is to stretch this out so it uses all of the available range.

Does some combination of aperture and shutter affect this? In my experience it simply shifts the curve more towards the dark end or the light end, but the curve itself remains the same width. In a studio setting, would lighting help to achieve this result? If so, how would you accomplish this outdoors?

If it matters, the camera in question is a Nikon D50. I know you can change the shape of the contrast curve in the camera, but my understanding is that this is simply done in processing, i.e. it does the same thing that photoshop would do, it just does it in the camera. But the same artifacts would be introduced. (But please correct me if I am wrong about this).
posted by Pastabagel to Media & Arts (26 answers total) 16 users marked this as a favorite
 
With the HDRI technique, you can take multiple pictures at different stops ("exposure levels"), and then run an algorithm to convolve them into one picture.
posted by Blazecock Pileon at 11:07 AM on December 12, 2007


Does some combination of aperture and shutter affect this?

No. You are right that just shifts the curve up or down. You will have more dynamic range if your values are high than if they are low, however. So if you're going to do any post processing, err on the side of overexposure (avoiding clipping still of course).
posted by aubilenon at 11:12 AM on December 12, 2007


Oh also if you're going to do a bunch of processing, shooting RAW will give better results, and more dynamic range (it's 12 bits per channel instead of 8)
posted by aubilenon at 11:15 AM on December 12, 2007 [1 favorite]


I don't think there's anything in terms of aperture/shutter settings that would significantly affect the situation. Some lenses are described as more "contrasty" than others (I don't have hard data to back this up), and using a lens hood can help with contrast as well, so that might be something to try.

In terms of lighting, in the example you gave you could place the main light at a shallow angle to the surface of the paper in order to exaggerate the shadows. Outdoors, a polarizing filter can really help with contrast.
posted by 0xFCAF at 11:17 AM on December 12, 2007


1 - use film, or
2 - underexpose and then lighten the image in photoshop
posted by caddis at 11:19 AM on December 12, 2007


Caddis has it. Use film. You'll have dynamic range coming out your ass.
posted by chunking express at 11:20 AM on December 12, 2007 [1 favorite]


No one's mentioned graduated neutral density filters yet? This is an easy way to rock landscape photos, bringing the sky down a few stops and exposing a larger range of the ground. This would work for non-sky, too, but pretty much only things you can line up with the ND border.

Supposedly there are ways to boost exposure of B&W (on film...maybe digital too) using color filters. I don't have any experience with that, though.

But yeah, B&W film has the best dynamic range of any single-shot capture possible. So much so that you actually need to compress the range further since photopaper doesn't have the same range.
posted by cowbellemoo at 11:38 AM on December 12, 2007


Use black and white film and learn the Zone System. Here's kind of a shortcut to the Zone system. Get a spot meter. Meter the shadows in your scene and place them on Zone III by underexposing 2 stops from the meter reading. Meter the highlights and make a note of the difference in stops. If it is 5 stops (inclusive of the high and low readings), use normal development. If it is fewer than 5 stops, expand development (higher temperature or leaving it in the developer longer) according to the chart for your specific film and developer (you can find this information in the datasheet from the film manufacturer, available on the web, or any number of independent charts). If it is greater than 5 stops, contract development. The reason this works is that the shadows are laid down first in development, but it takes longer for the highlights to get laid down.

You can apply some of the lessons of metering to digital photography, for instance, avoiding blown highlights by underexposing, and then using curves to raise the shadows up - but you cannot change the dynamic sensitivity of a CCD sensor. If you insist on sticking with digital, your best bet is to use a tripod, take the exact same shot with different exposures (use the built-in spot meter to determine the range between the highlights and shadows of your scene), and then merge the shots using any number of available software (I don't know the names of the software since I don't use them - I absolutely hate the fake-looking HDRs produced by people who don't understand the principles of exposure and are just pushing buttons).

On preview, color filters don't intrinsically alter the exposure - they just affect how much light gets through to the film at different wavelengths - for instance, an orange filter blocks out more blue light (the opposite of orange), causing a blue sky to register as thinner on the negative and darker on the print.
posted by matildaben at 11:43 AM on December 12, 2007 [2 favorites]


You will be capturing more data if you shift the curve to the right as much as you can without overexposing. The math behind this is kind of complicated. But basically, the range of brightness values are not evenly distributed from black to white--there are many more gradations as you get to the lighter side. Shifting the curve to the right captures more data in this range so when you work with it later, your image processing software has to do less "guessing."
posted by Brian James at 11:47 AM on December 12, 2007


Thank you all for the great (and speedy) answers. Does anyone have any explanation of why film has higher contrast? Do different films at a given speed have different contrast curve characteristics? (E.g. On a test image, Velvia results in curve X, Kodak's E100VS results in curve Y, etc.) Has anyone measured this and documented the results?

I have noticed that my 50mm/1.4 gives better contrast than the 28-80mm kit lens at 50mm, but it is considerably slower, so I had dismissed this. Perhaps the glass is simply better...

Do higher megapixel cameras (with same lenses, aperture and exposure settings) produce more contrast?

Do canon or nikons have inherently and significantly different contrast characteristics?
posted by Pastabagel at 11:55 AM on December 12, 2007


Yeah, you can't.

Think of your example of crumpled white paper - you've got 5 or 6 stops of dynamic range on your sensor. However, you just don't have 5 or 6 stops of light difference in the image. Depending on how you're lighting said piece of paper, you've got a lot of high white values (close to all 1s), and a lot of medium white values, then probably some middle grey in the shadows of the paper. There just isn't going to be any dark shadow in the exposure. As you move up and down your exposure scale, you're just going to get the same histogram, shifted left or right. There are ways to increase contrast on that scene, like lighting differently, as has been said. However, there's only so much you can do with that.

Here's the thing - why are you trying to do this? What's the point? Not all images are going to have the tonal range that leads to high dynamic range.

As far as artifacts are concerned, I'm not sure what you're talking about. If you're using levels and curves appropriately, on large images, you shouldn't really be seeing artifacts. Explain a little more what you mean, maybe? Is there a specific reason you're trying to avoid post-processing?

On preview, only print film has higher contrast - slide film generally has 3-4 stops, which is why it's so damn unforgiving, and most people bracket when shooting it.
posted by god hates math at 11:59 AM on December 12, 2007


film has higher contrast because digital formats (except for RAW I think? not sure about this one) are limited to 256 levels of contrast per color, where film, being an analog format, doesn't rely on a presentation method that requires printing ink onto paper or lighting up electronic pixels. Of course, unless you're using EXCEPTIONALLY expensive equipment (or just doing traditional photo development) you're going to hit that 256 level limit if you scan a photo or negative anyway. in a digital format, do what blazecock pileon suggested. Your camera should actually be able to shoot a range of exposures automatically with one press of the shutter button, I believe.

go here for an interesting explanation of the dynamic range limitations of digital photography and tips for getting around it using your camera and hdr image convolving software.

apologies in advance for any poor wording in my explanation. I understand the stuff but don't quite have the terminology perfect.
posted by shmegegge at 12:06 PM on December 12, 2007


(That should not have said 3-4 stops of contrast - clearly I meant dynamic range)
posted by god hates math at 12:06 PM on December 12, 2007


Do different films at a given speed have different contrast curve characteristics?

Yes. The manufacturers release data sheets like this one from FujiFilm that plot the curve characteristics.
posted by xo at 12:15 PM on December 12, 2007


Thank you all for the great (and speedy) answers. Does anyone have any explanation of why film has higher contrast?

There's a lot of confusion between dynamic range and contrast going on here.

People have been mentioning the dynamic range of film, not its contrast. If anything, the cases you've described are the opposite problem - you have plenty of dynamic range (as indicated by the fact that your exposure lands in the middle of the histogram with no clipping on either end), but not enough contrast. While the dynamic range of a medium determines the maximum contrast it can describe, it doesn't necessarily mean anything in terms of how well it expresses small differences in exposure.
posted by 0xFCAF at 12:16 PM on December 12, 2007


Here's the thing - why are you trying to do this? What's the point? Not all images are going to have the tonal range that leads to high dynamic range

This is exactly the point. I will of course process in photoshop eventually to color correct, etc, I just want to do all I can to take the best photo possible before I get to the processing stage.

In a natural setting (i.e. outdoors where I cannot control the lighting) there are circumstances where you want to enhance the contrast (think of trying to photograph clouds that have exceedingly narrow tonal range). My goal is to get as much dynamic range as possible "optically" before I post process, because any post-processing creates artifacts (little pixellations that in my OCD I can't help but notice and obsess over).

Now, clouds are unique in that they refract a considerable amount of light that they can become luminous in their own right, and such that a polarizer can help dramatically to isolate clouds from the sky and one cloud from another (clouds that are at different altitudes but which appear to be on top of each other from the ground), so this is why I didn't use this as an example.

I guess the answer is that it can't really be done in digital, but now I am very much intrigued about the possibility of using HDR to accomplish this, as that will at least provide image data for every part of the range across all of the shots in the bracket. I simply assumed that all HDR was good for was creating those annoying fake-looking images that are all over flickr, but BP's and matildaben's comments are making me wonder if this effect has its uses when applied delicately.

So so far the way to get the best image before post-processing appears to be to use a filter when appropriate, shift the curve to the right as much as possible without blowing highlights, shoot in raw, and experiment with hdr.
posted by Pastabagel at 12:32 PM on December 12, 2007


god hates math and 0xFCAF have it; the problem you're describing is not that your camera doesn't have the dynamic range you need. The problem is that the scene itself doesn't have the dynamic range.

Example: If you take a photo of a grey sheet of paper on top of a darker grey sheet of paper and light it evenly, the dark grey may only be 1 or 2 stops darker. It doesn't matter whether you're using film or digital because either one has enough range to capture the image (no clipping on either side in the histogram).

You're correct in thinking that aperture and shutter speed aren't going to affect the contrast, they'll just shift the histogram up or down. As long as it's not clipping you won't gain much contrast either way. I say "much" because cameras aren't perfect and might compress some of the shadows if you underexpose the photo, even if it's not clipped.

To really do what you want, you need to change the lighting. In the studio this isn't too difficult; you can adjust individual lights so that the light areas are brighter and the dark areas are the same, thus increasing the overall contrast.

On location this could be tougher. You could recompose to include some darker shadows and/or brighter highlights, or you could keep the composition the same and use shades or gobos to block some of the light sources and darken the shadows (just make sure they're not lighting the entire scene, or you'll just be decreasing the overall brightness again).

Another alternative is to use flash to augment the natural light of a scene. If you can get your flash off the camera you can use it to brighten select areas which could create more contrast. Strobist is a great place to start learning about that kind of stuff.

HDR could be an option, though it's usually used for the opposite situation where you have too much contrast to record all at once. Most people will just use curves if you need to increase the contrast.

Finally, ask yourself if you really need the extra contrast in-camera. Yes, adjusting the curves might create imperfections but are they really noticeable (especially when looking at the image at a reasonable size and not 100%)?
posted by sjl7678 at 12:46 PM on December 12, 2007 [1 favorite]


In the example given above, simply increase the contrast settings of the camera. This will result in a histogram that uses more of the available range.
posted by BobsterLobster at 12:47 PM on December 12, 2007


There's a lot of confusion between dynamic range and contrast going on here.

No kidding. This thread is full of misinformation.

Pastabagel, the best way to do what you want to do is to nail the exposure. Your sensor has a limited dynamic range, and what you want to do is capture the information available in the scene without clipping the pixels at the low or high end. At the bottom end of your sensor's response curve, the signal is limited by noise (called dark current). At the top end, it is limited by the way silicon goes "nonlinear" after the sensor sites are saturated.

So what you want to do is land your pixels right in the middle of your sensor's range, or a little to the right on the histogram. That's what you do when you nail the exposure of the scene. Matrix metering is often the most successful method in my experience. Then, you want to use your camera's raw format. In my case, with a Nikon D200, the raw format has 12 bits of "intensity resolution". Now, a JPG from the same camera has only 8, meaning that things will be posterized at the top and bottom of the response curve. So nail the exposure, and use raw to maximize the intensity resolution available to you in post.

Now, your example with crumpled paper is a bad one, because a crumpled sheet of paper does not have a lot of dynamic range, just like an overcast scene does not have a lot of dynamic range. That means that if you nail the exposure, you will have a little spike in the center of your histogram. However, you will be able to maximize the information that is there and resolve the fine gradients that you'd like to resolve. Capturing in high dynamic range, using filters, etc is meaningless in this case because the subject DOES NOT EXHIBIT high dynamic range.

Now, in the case of clouds, you may be faced with a very high *intensity* light source, which will exceed the ability of your sensor to respond to light if you expose for the landscape below. In this case, to capture the intensity range present in the scene, you may wish to use a bracketed HDR capture as mentioned above. You do not need an ND filter if you capture two shots, one exposing for the clouds and one for the landscape. You can simply blend them in photoshop after the fact.

Regarding lenses, "lens contrast" is a function of that lens's modulation transfer function, which is well-explained in this article.

I don't agree with all the suggestions to "use film" in this thread. Modern sensors have about 11 stops of DR available, which is absolutely awesome, and it's cheap and easy to bracket. Sensor performance is largely oustripping film. What's silly is that the reason film has so much "dynamic range" is that it nonlinearly compresses very intense stuff, like clouds. From what I've read in your question, you want the opposite effect, lots of resolution across the intensity range of your subject so that you can make fine gradations.

Nikon DSLRs do some processing before saving raw, and make suggestions in the EXIF data. Be sure to ignore these suggestions and see what you can extract using just Adobe Camera Raw or whatever. You might try changing the contrast curve to see if it affects your raw data, but I suspect it only applies to JPGs.
posted by fake at 1:07 PM on December 12, 2007 [4 favorites]


sjl7678 also has a great answer.
posted by fake at 1:08 PM on December 12, 2007


Thank you fake for your great answer.
posted by Pastabagel at 3:56 PM on December 12, 2007


I'm glad you found it helpful. If you get any results that you are happy with, and put them online somewhere, I would love to see them.
posted by fake at 5:47 PM on December 12, 2007


The only way to do it is to change the light.
posted by bradbane at 2:34 AM on December 13, 2007


You're correct in thinking that aperture and shutter speed aren't going to affect the contrast, they'll just shift the histogram up or down.

I disagree. Over and underexposure dramatically affect contrast.
posted by caddis at 5:03 AM on December 13, 2007


You will have more dynamic range if your values are high than if they are low, however. So if you're going to do any post processing, err on the side of overexposure

This is correct for film, but the opposite is true for digital. It's obviously best to "nail the exposure," but with digital, err a bit toward underexposure if you're not confident where the nail hole is.
posted by gum at 9:59 AM on December 13, 2007


You're correct in thinking that aperture and shutter speed aren't going to affect the contrast, they'll just shift the histogram up or down.
posted by sjl7678 at 3:46 PM on December 12, 2007


You are correct about shutter speed, but aperture impacts the MTF of the system, which describes contrast versus spatial frequency (essentially, contrast based on the dimensions of the details you are imaging).

caddis is also correct (poor exposures limit available contrast), but that is probably obvious.
posted by IAmBroom at 5:04 PM on November 30, 2008


« Older Freeware for Alarming Librarians   |   Derive time of conception or birth from DNA... Newer »
This thread is closed to new comments.