lossless
- Lossless will be shown at pixilerations 09/28 – 10/14/2007, Providence, RI
- Lossless has been accepted into collision11 and the Boston Cyberarts Festival.
- preview at Harvard Film Archive, 4/2007
- preview at UCLA Graduate School of Education & Information Studies’ workshop on forgetting and exclusion
description
We’ve seen “The Wizard of Oz” more times than any other movie. We may imagine it playing on a big screen in full Technicolor as it was shown in 1939. But the truth is that we’ve only seen it on television. Now that we may buy the digitally enhanced DVD and watch the “Wizard of Oz” on our laptop computers, we wanted to know exactly how the media had changed. By capturing the differences between a 35mm print and a digital version, Lossless shows exactly what has changed.
We started the project by looking for the best and worst versions of “The Wizard of Oz.” We soon realized that it would be impossible to secure the best version. George Eastman House has preserved an original negative (or negatives as it is a three-strip technicolor separation print). These negatives were scanned to create the 2005 Ultra-Resolution DVD. This DVD looks better than any other release we have found, including the 35mm print. So we were led away from the idea of measuring against a best version to calculating the total difference between two releases of the movie.
Media re-issues always promote improvements in fidelity. This project asks you to consider that fidelity by showing the difference between the film and digital renderings of “The Wizard of Oz.” What is greater fidelity for those of us who grew up watching this film on Television? It is curious to see detail in the shadows, to see a rivet in the middle of the Tin-Man’s face, and to see Kansas in sepia tones for the first time. What we really want is to get back in touch with that feeling we had the first time we saw the film. This feeling is something that cannot be restored by fidelity to the original, at least not for the TV generation.
Differences in media may be measured and these differences captured. Understanding the difference is another way to restore fidelity. Lossless means to clarify that difference.
compression and artifacts
The compression ratio for DVDs from frame scan to final disk is approximately 400:1. About a third of the compression is done with lossless codecs retaining all of the color, resolution, and detail captured in the original scan. The remaining two thirds of the compression is done with lossy methods which replace the original with areas of flat color and good guesswork. The image quality of DVDs is a testament to the skill of the engineers and psychologists who designed the encoding algorithms. The pertinent branch of psychology is known as psycho-physics. It is concerned with the range of sights and sounds that our perceptual systems afford us. For example our hearing is a curve that starts at 20Hz., peaks at about 4000Hz, and drops off at 20,000Hz. So the sound will be tailored to match this curve. There’s no sense in reproducing sounds above or below the curve, and this saves space when it comes time to compress data. I have used sound as an example, but in fact sound data typically accounts for only about 3% of this information on the disk. The rest is picture.
After time sampling (24, 25 or 30 frames per second) the picture is divided into a matrix of tiny dots called pixels. Each pixel is described with one luminance (brightness) value and two chrominance (color) values (4:4:4). The chrominance is typically thinned out to 4:2:2 (removing half the original color) or even 4:2:0 (removing three-quarters of the original color). After the color has been thinned the picture is broken up into 8×8 pixel blocks. These blocks are analyzed for redundancy in luminance and color in three dimensions: per frame (x,y) and over time (z).
Because video is picture (and sound) information transmitted by signal, it is possible to process the media with cosine transforms. The signal is treated as a sum of sinusoids with different frequencies and amplitudes. These lossy calculations (variations on Fourier transforms) are the fastest way to compress the data stream.
Artifacts provide the signature for every medium. We were always aware of the qualitative difference between film and video. Film draws one frame 24 times per second. Video draws half of the picture approximately 60 times per second, depending on the persistence of phosphors to knit together the whole frame about 30 times per second.
calculating the difference
We started the project by looking for the best and worst versions of The Wizard of Oz. We soon realized that it would be impossible to secure the best version. George Eastman House has preserved an original negative (or negatives as it is a three-strip technicolor print on flammable nitrocellulose stock). These negatives were scanned to create the 2005 Ultra-Resolution DVD, and ironically this DVD looks better than any other release we have found. So we were led away from the idea of measuring against a standard to calculating the total difference between two more releases.
Frame 40 from the Ultra-Resolution DVD, 35mm print scan, MGM laserdisc, and the resulting difference. [click for detail]
First the media is rendered into a common denominator. For lossless we scanned the 48 frames from a 35mm film print at 2k resolution (Academy 2k is 1828×1332 pixels => 2,434,896 total total per frame) and up-rezed the PAL DVD (resolution increased using nearest-neighbor interpolation) to match. Whereever the pixels were the same we paint them black. When they differe we average the luminance and color of the difference.
installation
This version of lossless utilizes a 16mm projector to show a loop of difference frames. Theaudio has been prepared in the same way.