Formalizing Photographic Aesthetics

by Philip Greenspun

Site Home : Research : One Article


These are some research ideas that I'd like to pursue but don't (because I keep working on integrating relational databases and the Web).

Better controls for photographers

What are the appropriate controls for a camera? Traditional cameras have controls that are artifacts of the physical process of silver-halide photography, e.g., aperture and shutter speed. Computerized cameras with their "portrait" and "action" modes pull away from these direct physical controls in a crude manner. In an age of digital imaging, it is worth asking if cameras could be built with direct aesthetic controls, e.g., a "romantic/clinical" knob or an "Ansel Adams/Gary Winogrand" slider.

Before heading out to take pictures, a serious photographer makes choices that have profound aesthetic consequences given the limitations of silver-halide processes. For example, if a 1950s style is desired, Kodak Tri-X black & white film will be selected for its distinctive grain and look. Even if shot under ideal conditions in a well-lit studio, the resulting images can never be used where color is needed or a grainless modern style is desired. Perhaps with appropriate digital imaging technology, the photographer could capture detailed information on fine-grain color film or with an all-digital camera, then process the images to look as though they'd been shot on Tri-X.

For a photograph of a tall building that preserves the architectural lines is required, a photographer is currently forced to use a view camera, i.e., a camera where the lens and film are not fixed parallel to each other. It may be possible to simulate some view camera movements with digital processing and thereby enable the photograph to be taken with a simple Nikon.

Achieving a grainless look and high color saturation in a traditional print requires using low-speed (e.g., ISO 25 or 100) film that may require compromises in depth-of-field, freezing action, and lens weight. If the final image is intended for distribution via the World Wide Web and for presentation on a computer screen, it may be possible to achieve indistinguishable results using ISO 400 or ISO 800 film via digital processing. It would be nice to have a formal method for answering a question of the form: What is the maximum speed film I can get away with given my desired output size?

Sheet-film photographers such as Ansel Adams affirmatively control the contrast of each image by varying the development time. This is not possible with roll film, e.g., 35mm, where the entire roll must be developed at once. However, by using wide-latitude film and making multiple exposures at different shutter speeds, it should be possible via digital processing to assemble a final image of any desired contrast with detail in the shadows and highlights limited only by the output medium. If the film or imaging sensor has the property of silver halide crystals that a threshold certain number of photons are needed before any image will be registered, it may be desirable to pre-flash the entire frame before trying to capture shadow detail. The question here: How many exposures and of what kind do I need with a static subject to completely capture the scene for all practical aethestic purposes?

Producing exhibitions

Why is it that a book like Manhole Covers (MIT Press) is art and a collection of snapshots, each containing a manhole, is not? If a principal reason that Manhole Covers is art is the consistency of presentation, i.e., same cropping, same proportions, same toning, same grain, then can we turn snapshots into art under computer control? Can we help the would-be artist by giving a collection a photos a series of "looks" and letting the artist choose the best one? (Note that in a photo exhibition, you usually see only one kind of photo, e.g., black and white prints of 16x20" in size, framed in silver aluminum, or vertical Cibachromes, 20x24" in size, all from Fuji Velvia originals.)


Photographs gratuitously inserted into text (in hopes that more people will read this) are copyright Philip Greenspun. You can find 3000 more in my photography and travel pages.


philg@mit.edu

Reader's Comments

First, I am amazed at the things you have done in the last three years. I have talked about you to dozens of people but it was pre-ArsDigita. A lot of water under the bridge in three years!! Obviously, your web experience is impressive and the photography is fantastic.

I shoot a Mamiya M645 when I am not chasing my 2, 6 and 8 year old boys around with my wife and when we get a chance to travel. Carried that sucker across Australia/New Zealand and the Middle East.

Sorry for the aside, but the point of my comment was...you state "If the film or imaging sensor has the property of silver halide crystals that a threshold certain number of photons are needed before any image will be registered, it may be desirable to pre-flash the entire frame before trying to capture shadow detail. The question here: How many exposures and of what kind do I need with a static subject to completely capture the scene for all practical aethestic purposes?"

With the recent phenomenal advances in digital capture rate in digital cameras [Example: (http://www.dpreview.com/news/0101/01010803twonewjvcs.asp)], do you have the question answered yet? Do the 3.3 megapixel cameras offer a "clean" palate for which all encoded images start from scratch?

Note - good luck with your lawsuit, and thanks for the incredible amount of writing across the website and links that I am relieved to see has NO misspellings! What a relief to see a website that is clean, funny, and written well! Steve Daugherty

-- Steve Daugherty, May 9, 2001

Dear Philip,

I like your essay on formalizing photographic aesthetics.

Imagine locating many different photographs in a multidimensional space, whose dimensions need to be figured out (character of the color, depth of field, composition, and 3 to 6 others). This space needs to characterize, gracefully and reasonably, a great variety of different photographic styles that we may seek at one time or another to emulate.

The task of the automatic-aesthetic camera (AAC) is to sample that space so that a reasonable number of points (photographs) are available for post-shoot processing. One picture cannot capture all your want. So you need smart bracketing to replace what is now rather naive bracketing. Thus the AAC will provide a set of wildly different pictures as it systematically moves through the multidimensional space of possible photographic styles--all applied to the same scene.

Now this is similar to what you are trying to do now with nouns (names of different photographers), as each noun in effect describes levels of 4 to 10 variables. So you could characterize 20 photographers by where they are located on the fundamental aesthetic variables. But my proposed method of sampling a multivariate space will produce combinations of aesthetic variables that no one has come up with or turned into a style.

This all would displace the act of making a photographic style from camera controls to the computer editing. Other than editing, the key action in making good photograph then becomes a matter of just showing up, which it already is. But with the AAC multiples to edit afterwards, you get a lot more information out of just showing up.

As ever,

ET



-- Edward Tufte, October 2, 2001

IEEE Computer, August 2006, is a special issue on computational photography. What can you do with GIGApixels? Merge different focus, contrast, and viewpoints to get high dynamic range, near-3D viewing, and "X-ray vision"! Not an "Ansel Adams" control, but shows how computation can help.

-- Paul E. Black, December 3, 2007
Add a comment | Add a link