Some Technical Matters

The photos presented here represent a sample of our archive of more than 2,400 retained photographic images we collected during our initial fieldwork in 1989-90. Other, more recent photographs may be included at a later date.

None of these images started out life as digital ones. Rather, they were taken by two film cameras, a Canon FX SLR with a range of professional quality lenses, and what was in those days a fairly high end Canon point and shoot with a zoom lens. 77% of the pictures taken generated color slides, 10% color negatives, and 13% black and white negatives.

Film, particularly color film, is notoriously sensitive to degradation by heat and humidity. Posh photographers working in the tropics accordingly try to keep exposed and unexposed film refrigerated, get it developed as soon as possible, and quickly remove it to more benign environments. There is probably no place on earth more perilous to film than Bangladesh in the summer, especially when flood drives the already high humidity through the roof. Ours had a hard time of it from the onset. As anthropologists we were not living the life of posh photographers, but rather in a single room without electricity. There was only one power line in the immediately vicinity, and to the best of our knowledge no refrigerators at all. For nine to twelve months our film was subjected to temperatures that stayed in the high 30°s C, and to humidity levels so high that mould grew inside camera lenses.

At the time we were thinking of photographs largely as a kind of ethnographic data, although earlier flirtations with art photography probably played a minor role in what we shot and how we composed things. We therefore used standard off the shelf films suitable for available light photography, chiefly 100 ASA (as it was in those days) Fuji slide and negative film, and 400 ASA Kodak Tri-X. We therefore also handed all our color materials to big box commercial outfits for development rather than send them to specialty labs where they might have received better treatment. With the few exceptions, black and white film was processed by us in our own darkroom, to a fairly high standard. After processing all these materials were kept under good conditions almost untouched for 17-18 years. Nevertheless, time is an enemy of all dye-based color films, and further degradation did occur over this period—a little in white balance and a lot in overall color balance, as some dyes faded disproportionately.

Our first goal in archiving this material for posterity was to digitize it in a way that would retain as much of the extant information as possible and would add the fewest artifacts and distortions. Very high quality scans were necessary to achieve this first step. Accordingly, we employed a professional quality film scanner, a Nikon Coolscan 5000 ED. This scanner has been widely seen as generating about the best scans of 35mm materials that can be achieved short of drum and flatbed scanners costing well over $10,000. This scanner proved up to the task, and desite it having been discontinued we would highly recommend it to anyone contemplating a similar project.

An issue that needed to be settled at the onset was at what resolution we should scan. Our tests confirm that the 5000 ED is capable of capturing useful detail at its maximum resolution of 4,000 dpi. So why not scan at that resolution? A single uncompressed 16-bit tiff file of a color slide scanned at 4,000 dpi weighs in at an astonishing 138 MB. Processing times increase proportionally. For us, though, the chief difficulty with scanning at this resolution is that archiving becomes problematic from a storage point of view. After consulting common practice we did a series of scans at increasing resolution, both with and without post-scan sharpening. We confirmed what others had found, that scans done at 3,100 dpi were indistinguishable from those done at 4,000 dpi at normal viewing sizes, and that a carefully sharpened 3,100 dpi image appeared as sharp as an unsharpened 4,000 dpi one at all magnifications. We therefore scanned all color materials at 3,100 dpi, producing uncompressed 16-bit TIFF files approximately 72 MB in size.  As they produced smaller files (40 MB or so), all black an white negatives were scanned in at the maximum scanning resolution of 4,000 dpi.

One might wonder why we chose 16-bit tiff format for our scanned images rather than something more economical of space like (8-bit) jpeg format. It did not take long to discover that many of these scanned images required a lot of corrective work due chiefly to film degradation, but also because of the remarkably varying nature of available light when the photographs were taken. We also wanted to archive images that incorporated as much as possible of the information on the film, so that others might have a full range of possibilities of manipulating them at a later date. 16-bit tiff files were definitely the way to go.

The total archive of image files takes up 150 GB, not including catalogue files and the like. While this is no longer too large to make available on the Internet—which it was when we began the project—downloading and viewing 72 MB image files is still not practical. Reduced from 16-bit to 8-bit and converted to ‘lossless’ jpeg files with post-scanning adjustments permanently included (and therefore no longer the original images, strictly speaking) the color image files go from 72 MB to 13 MB and change—still not providing a realistic prospect for Internet archiving with time effective downloading. ‘60% quality’ jpeg compression reduces this further to 1.6 MB, with some definite degradation of quality at high magnifications but rather little at screen size. By scaling 60% jpeg images down to a maximum size approximating a full computer screen image (1,350 x 900 pixels) we further reduce the typical file size to 300 kB. In either of the two last mentioned formats we could archive all 2,400 images on the Internet right now. There are both positives and negatives to this option, and we will be considering it carefully in the future.

A second major goal in the development of this archive was to restore the images to their original state. Perhaps it is more accurate to say that our objective here was to as closely as possible reflect the look an feel of ‘being there’—without question an interpretive process as well as a technical one. To begin this process we did everything we could to ensure than scanning itself did not introduce arbitrary or irreversible corrections to the scanned images. Like most of its peers, the Coolscan 5000 ED is controlled either by proprietary (Nikon) or optionally purchased scanning software like Vuescan. In the end, we used Nikon Scan 4.0, which came with the scanner. Such software is capable of ‘correcting’ the scanned image in a number of ways, post-scanning. If such options are used each initial image will be altered in unique, permanent, irreproducible ways. Moreover, while these kinds of software corrections are often surprisingly effective, Bangladesh available light photography  must contend with an enormous range of brightness, white balance and color saturation that can often fool the best intentions of software. Instead, we followed the advice of C F Systems and did flat, linear, uncorrected 16-bit scans with the scanner set to a gamma of 1.0. The scanner's automatic exposure option was turned off, but auto focusing was maintained. Color management was not used during scanning. The only other scanning software option used was Kodak's Digital ICE 4.0, a non-invasive dust removal scheme used on color materials. Color and black and white negatives were scanned as positives, in order that the scanner software not introduce inaccuracies in inverting the negatives to positives.

The resulting ‘raw’ color images were just dreadful to look at, chiefly as a result of the selective degradation of some of the dyes. Here is a particularly bad example of a scanned slide:

Also, while we had found them just fine for the production of ‘wet’ darkroom prints, ‘raw’ scanned images of our black and white negatives were far too contrasty, with blown out highlights and muddy shadows.

Here is an example of a somewhat less degraded slide scan scan looked like as it came from the scanner, adjusted only for exposure. In this example, the initial color balance is much better than most other slides (certainly better than the above example), but the contrast, saturation, highlights and shadows all are greatly off:

Raw scan of old damaged slide before adjustment

The repair of many slides required fairly heroic measures. All images were brought into Photoshop CS3 or CS5 (using the PhotoPro color space), where they were filtered by either C F System’s ColorPos or ColorNeg plug-in. These proved to be the lifesavers they are intended to be, based as they are on sophisticated algorithms to restore, or at least ballpark, color balance (among other things). ColorNeg also was used to invert the negative images. This is the same 'raw' slide image after it has come through C F Systems ColorPos alone:

Old damaged slide after filtered by CF systems ColorPos

ColorPro has since been replaced by ColorPerfect, which has been used more recently.

Some images were then roughly adjusted for brightness, contrast and shadows in Photoshop, but most were not. Next, all images were imported into Lightroom 2.x; we are now using Lightroom 4.x. Lightroom allows powerful cataloguing options. It also allows us to do a range of reversible development and cropping changes on each image. Initial cropping for archival purposes was the minimum needed to clean up the margins of scans. The goal of these developing adjustments was to match our sense of what things looked like rather than slavishly restore the images to their original state. For comparison, here is the Lightroom ‘corrected’ version of the slide image immediately above.

Damaged slide after final adjustments

Here is another before and after example, in this case of adjusting for strong color shift, chiefly towards the blue. The first 'raw' image was adjusted only for exposure:

Old damaged slide of boat, Bangladesh, uncorrected for color


Old damaged slide of boat, Bangladesh, corrected for color

And what of the first example above? Here it is after all adjustments. Not great, but a lot better:

Raw scan of old damaged slide before adjustment

In some cases the differential fading of color dyes has been extreme. In this case (a negative), the inverse of the red dye persisted while the others faded. Looking at the photo below you probably would not know that the man in the foreground has colored his skin magenta:

Old damaged slide with strong red shift

With the reds reduced and some other color adjustments:

Old damaged slide with strong red shift corrected

With final corrections:

Old damaged slide fully corrected

We acknowledge that memory is a slippery and unstable thing and that we may have remembered things wrongly. Moreover, in making adjustments we were not perfectionists: the sheer time required made this impossible. The Lightroom changes we have made are reversible, and we have a wide range of digital images of Kazipur taken in 2008; in the future we may use these to do further corrections on the photographic images taken in 1989-1990.

It was a tough choice, but in the end all image adjustments were made with a color corrected monitor: initially a Dell 2001FP and more recently a highly recommended Dell U3011. We used Datacolor's Spyder Pro/Elite 3.0 or 4.0 to adjust the monitors. One might ask why color correction was a tough choice. In the master archive we are aiming at a kind of veracity, so this was the obvious thing to do. Otherwise we would be making corrections based on an irreproducible standard. Web presentation poses a different challenge, because because in practice there is no standard: few people will be viewing these Web-based images with a color corrected monitor. Uncorrected monitors tend to be overly bright and over saturated, with high gamma and skewed white balance, but none of these is consistently ‘off’. Moreover, the world's most popular browser, Internet Explorer, is still not color managed in its earlier incarnations, even in IE8. While IE9 and IE10 are color managed, we strongly recommend the newer versions of Firefox, which are very inteligently color managed.

All Bangladesh images found throughout this site were managed in the PhotoPro color space (applied first in CS3) and exported from Lightroom in jpeg format, using the sRGB color space. Thematic photo essays were organized in Lightroom and the Flash versions exported through the highly recommended Slideshow Pro Lightroom plugin. Slideshow Pro also generated the Flash materials through which the essays and associated slide shows are presented on the Web. Adobe Flash Player is required to view the recommended version of the thematic photo essays. As some folk do not have Flash installed, we have made a parallel set of pure HTML slideshows using the Impact Lightroom plugin.