Goethe sought his urforms, Plato had his solids. Today, we owe our digital stimuli to a most peculiar collection of ancestral forms, an eclectic assortment of objects and images worthy of Wilkins. Our contemporary collection includes:
Extending the bounds of this group slightly, we might also include the full texts of As You Like It, Paradise Lost, and Alice in Wonderland. These items are among the core standard test objects for many of the computing processes we now take for granted. Their function lies somewhere between that of a crash test dummy and Tippy the Turtle ("Can you draw me?). Early in the development of encoding, transmission, and rendering processes for images, sounds and other files, researchers landed on these mostly arbitrary objects as their common test subjects. Each object has been compressed and re-compressed across multiple algorithms and analyses on the way to establishing public standards, licensed versions, and proprietary protocols. In other words, almost all of the digital images, texts and sounds we view each day are based on processes first tested on these objects. Only when these objects came into clear focus with each new algorithm were the rest of us allowed to add a new capacity to our expanding digital sensorium. At their least interesting, these objects functioned simply as test patterns, or "Hello World" messages. One can't say that every jpeg today bears the mark of "Lena" the Playboy Model. Yet researchers depend on experience of these objects to make subjective judgments about which human physiological processes to facilitate or ignore. Lena isn't exactly the "face that launched a thousand ships" - we'd still have jpegs without her image. However, the way in which her image has been reduced repeatedly to code represents something of a prototype class, a way of seeing intended to be repeated over and over again. With this post I'll launch a survey of this Wunderkammer, with some words on the origins, rationale and purpose of each object. Many of these are steeped in lore and legend. My information, gleaned from internet research and the ACM archive, may be incomplete. This collection may not grow much beyond its current confines, as computing problems seem to require an increasingly diverse set of increasingly proprietary test data. In a later concluding post, I'll reflect a little on these changes, and their implications for future digital sensoria. There is some great critical work going on in software studies and infrastructure studies that can help us understand these processes. For today, we'll start with the obvious, the "lena.tif" file.
[Cropped image of Lena Söderberg / source: Playboy, November 1972] Probably the most famous test image, this over-the-shoulder pose is certainly the most politically charged of our set. The image is of course a cropped version of a 1972 Playboy centerfold image. A website at Carnegie Mellon documents the history of this scan, with the help of a report written by my colleague Jamie Hutchinson. (Follow the links to some especially bizarre documents of the model attending a recent professional conference to celebrate the anniversary of her role. There is also a curious Woody Allen connection here which I challenge you to discover.) This image continues to be used today as a standard for establishing effective algorithms for sampling and interpolating images. In fact, once the origins of the image were uncovered in the nineties, Playboy relinquished some intellectual property control and offered access to the original negative, for increased resolution in testing. There are some obvious lines of interpretation begging for attention here. For example, it seems only fitting to base image compression formats on interpolation of a pornographic image, given pornography's prominent role in digital image circulation. The male gaze here has been encoded and streamlined into a non-optical information stream. (Some would argue that this process started long before software; see Harun Farocki's film Ein Bild (An Image).) The image has at least one obvious function within the labor of graphics research, attempting to provide heterosexual male workers with some degree of pleasure through the drudgery of repetitive coding. More than that, the image adds an extra-visual register to the subjective decision about how "lossy" is too lossy in processes of sampling and interpolation. Decisions on the limits of acceptable degradation in compression were a group effort, and assessment of perceived sex appeal could replace other, more contentious levels of tolerance for noise. (e.g., "She's lossy, but I think we can all agree she's still hot.") By contrast, one could imagine more strident arguments over acceptable lossiness in the case of a scan of the Mona Lisa. Fine Art is more quickly relegated to the domain of "personal taste," whereas the Playboy centerfold was constructed to create massive, objective appeal. "Our main aim was to compress the image without affect on Lena’s beauty," writes one contemporary researcher. We of course see here an implied ideal audience - or at least a demographic portrait of who gave us our more popular compression formats. In this way, other less popular test objects in image compression seem more like test patterns, and less like subjective rubicons. An image of a Mandrill has been a popular source, for example, as has been a cropped photograph of red and green peppers, or even a photograph of a street in Dorset. More on that next time. View the second installment of this series here. |
|||
Vague TerrainDigital Art / Culture / Technology |
||
|
|
|
