General thoughts on quality and technical considerations

While "digitization" sounds very computer-y and, well, digital, it is in reality a very analog, very human, hands-on process with little automation possible. As a team we produce tens of thousands of digitized images every year, each bringing our own unique backgrounds, talents, and sensibilities to bear.

All of the technical decisions involved in digitization are made using norms and best practices developed by the cultural heritage imaging community at institutions like ours all around the country and the world. Even with today's technology, though, we still must make tradeoffs.

It's worth emphasizing that preservation-level accuracy isn't required to provide online access to materials. We often face a choice between creating images that are less-than-perfect or not imaging the item at all - and making these archival materials digitally accessible always wins that argument.

There have been incredibly significant advances in the image quality achievable by DPG and similar groups at other cultural heritage institutions in recent years. You'll see evidence of both current and past imaging standards in the items included in this exhibit, some of which were digitized quite some time ago when imaging technology was nowhere near as good as it is now.

As image quality is something that continually improves with better equipment, resources, and training, our time and effort is usually best spent making new materials accessible rather than re-imaging past materials that were shot with “legacy” standards.

Where are the color bars?

You may be familiar with archival photographs from other institutions that feature prominent color bars. Color bars or other color targets are a critical part of the cultural heritage imaging process, though they may or may not be of use to end-users viewing the final images. There are different brands and types of color bars, and different ways of using them, such as an “object level” color bar which sits next to the object or “device level” color targets which are used for testing but do not get included in final images.

DPG performs device-level color calibration for each imaging session, which involves capturing images of physical color targets that get checked with calibration software to ensure that colors and tones are accurate (it even gives a “pass” or “fail”).

Despite this scientific calibration, we will still visually evaluate our images as there may still be slight variations in tone or color which the human eye is extraordinarily good at detecting. Additional evaluations are made during quality assurance checks, such as for camera focus, organics (something on the object like a hair or bit of fluff), or other issues that may need to be corrected with a reshoot.

Lighting

For our overhead cameras, where the bulk of our digitization takes place, the lighting setups have been designed according to cultural heritage imaging norms to create even lighting that illuminates as much of the object as possible.

Lighting is tested and calibrated daily to ensure that it is even across the imaging area and the object, while still allowing some shadows - shadows help to provide some visual depth which better communicates the three-dimensional characteristics of the objects.

Shadows help to demonstrate things like deep folds in stiff parchment, the texture of coarse handmade paper (and even in finer paper), or the relief designs on a wax seal. Carefully managed shadows are an important aid to our ability to represent the physicality of the objects.

Resolution

The details in any object may be measured using a magnifying loupe or other methods in order to identify the resolution necessary to clearly represent it in a photograph, a method called “item-driven image fidelity” (IDIF). Read about this method in an interesting article from the Smithsonian Institution Digitization Program here (PDF).

Based on extensive testing using that method, and in line with the Smithsonian's findings, DPG defaults to 600 ppi resolution - which actually far exceeds the needs of the majority of objects. This resolution captures fine details - including physical characteristics like the texture of paper - and is very manageable with our equipment for most items.

PPI stands for pixels per inch, meaning, in the case of 600 ppi, one inch measured on the original physical object is represented by 600 digital pixels.

There are cases where the details in the objects, fluctuating sizes of the materials within a collection, specific end-uses of the images, or other factors might prompt us to image them at 400 ppi in order to have a streamlined and efficient workflow. We typically don't go lower in resolution than that.

There are also cases where higher resolutions than 600 may be desired (or specifically indicated by the IDIF method). Photographic prints are usually imaged at 800-1000 ppi or more, along with items like artwork, coins, and other items which have fine details that benefit from a higher resolution.

Photographic film is a whole other realm, reaching as high as 10,000 ppi, with 7900 ppi being the standard for regular 35mm film.

Detailed notes on film digitization - resolution & other considerations

As with any type of material DPG works on, we are much more cautious with physical handling than is generally deemed necessary in a commercial setting. To that end, the film digitization method used in cultural heritage imaging does not use fluid mounting or other techniques employed as standard in high-end commercial scanning. The film is touched only by gloved hands, clean and dry anti-newton's-ring-glass film holders with no edges, and filtered air from a blower.

Dedicated high-end film scanners used commercially such as drum scanners and Hasselblad/Imacon Flextight scanners are no longer made, in any case, and meanwhile newer imaging technologies have advanced so significantly that the results we can achieve with our much more cautious methods already exceed scans from those devices, and with much faster throughput, with only minor drawbacks.

While resolution and tonality in our images exceeds typical drum scanning, the dust, debris, hairs/fibers, and so on that are very common in archival film collections are visually reduced by fluid mounting in a drum scanner and will unavoidably be more evident with our dry method.

We look for debris that covers important details or that would be difficult to remove with digital post-processing in our digitized images, and attempt to use a filtered blower to carefully move or remove the debris before imaging (multiple times, if necessary). We don't perform any type of digital cleanup for images that get accessioned into the Stanford Digital Repository, but the idea is to create images that would be easy for someone else who wishes to use them for publication or other reproduction to touch up as desired themselves.

The 7900 ppi number mentioned earlier as standard for 35mm film imaging is not arbitrary; it is what we can image in a single shot with our current equipment. It also happens to far exceed typical film digitization resolution standards (such as the FADGI standards often referenced in cultural heritage imaging).

For most practical purposes besides extreme enlargements resolutions this high are not strictly necessary, but we feel they are justified by the IDIF philosophy. Though the actual resolving power of film - not to mention camera lenses - is not actually that high, we feel this high resolution captures the analog quality of film grain similar to what is achievable with a very fine darkroom print. The analog tonality of the film is better captured than at typical resolutions, and aliasing and other things that give digitized analog film images a digital look are completely avoided, producing as analog-looking of an image that is possible while also being practical.

For medium and large format photographs, single-shot images must be done at lower resolutions - but our single-shot resolution for these formats also exceeds common imaging standards for medium format (e.g. 120 film and similar formats) and large format such as 4x5" and 8x10" film. Less resolution is generally necessary because the amount of enlargement required is significantly less, but we must still choose resolutions carefully to avoid a digital appearance to the images.

Stitching multiple shots together to obtain higher resolution images of larger formats is possible, and we make the decision to do so based on each particular piece of film. It is a time-consuming and sometimes difficult process with film compared to other types of materials that we commonly stitch such as maps and other oversized items, but it is justified in many cases.