The engraver's insight

Remarks written on returning from my first visit to Albrecht Dürer's house (or to Europe for that matter)
David Dailey, September 2008.

close-up of 300 dpi scan of Durer
Albrecht Durer, painter, engraver, Renaissance man.
Self-portait, (1500).
Engraved, apparently by Paul Krey, Leipzig (circa 1885);
cropped from the original scan (2008). [1]

a) Nuremberg was fascinating. If such a town were in the United States, it would be overrun with tourists, Disney would have to buy it, and admission for a family of four would be a month's salary.

b) I had always found Dürer's work impressive (a low quality reproduction of his Four Horsemen was thumb-tacked to my wall in my office in our old building at SRU for several years). I knew he was respected within the history of art, but a tour of his house taught me some of the reasons for that respect. He was a scientist, philosopher, chemist, successful businessman, artist and thinker. He also held opinions on copyright.[2] His thoughts and works influenced both Southern and Northern Europe.

Europe at the time was undergoing a variety of transitions: political and religious upheaval, awakening knowledge of the New World, plagues, explosions in fields of thought, science and the arts -- all enabled, to some extent, by advances in optics, metallurgy, and the technology of navigation. Dürer was well-positioned culturally, geographically, and financially to explore some of these advances. His insights on the relevance of it all were uniquely his own.

c) Fast forward 500 years to the present. What value does engraving have now that photography has arrived (some 150 years ago now)? Well, when I first started putting public domain images (like these) on the internet some 20 years ago, I discovered that engravings were special. They were single-bit images, for which the shading (half-toning in the more modern world of printing circa 1988), had been provided by a master craftsman. They were easily scanned, and the storage footprint they subtended on the limited server space of the late 1980's was small. Even at fairly low levels of spatial resolution (like 300 or even 150 dpi) a four bit greyscale scan preserves most of the engraver's markings (see here for some technical advice).[3] Furthermore, from the engraving we might, with a simple combination of filters, convert the engraving back to something resembling a photograph (see below, at right).

More importantly than some of these coincidental 21st century uses, however, is that the engraver provided insights into the structure and semantics of the image that the camera (because of its lack of a brain) simply does not see. The engraver's marks tell us something about where two regions (similar in greyscale values, but different in meaning) actually meet at a boundary. It signals, in some cases, using the directionality of hatch and stroke marks,  where depth and 3D curvature differ from shade and pigment. This information  is lost to auto-trace algorithms (regardless of their mathematical and algorithmic sophistication). In the engraving, the folds of Mr. Webster's scarf are distinguishable since the engraver's strokes head in different directions. The greyscale/photographic version obscures this detail that was imparted to the image by the insight of the engraver.

Let me try to illustrate this point.

Here we have, left to right, versions of the original, cropped to the face:

photo of painting of Durer greyscale of photo of painting engraving -- reduced in size

Photo of Painting (from Wikimedia and the Yorck Project) [4]

Same as at left in greyscale.

Shrinking of engraving.by Krey.
(Click to see larger version showing engraver's marks.)

The engraving, at right above, tends to (at this resolution) show greater detail than the photograph. A part of this, I think, is due to the engraver's eye (heightening detail which, over time -- 100 years has elapsed between the time the photograph was made and the time the engraving was made--  has become harder to see), and a part is an artifact of the low spatial resolution available to computer monitors.

However, as we zoom in a bit, and attempt a reduction of the image to lower chroma (such as would be found in a vector-based description of the original), we may observe some interesting phenomena.

detail of eyes in photo detail of eyes in engraving
Dürer'eyes photo -- 256 shades of grey Dürer'Eyes photo -- posterized -- 4 shades of grey Dürer'Eyes engraving -- 4 shades of grey

While a part of the greyscale information present in the photograph, does convey meaningful information when converted to a vector graphic (using conventional techniques of blurring and posterizing -- if you have Opera or IE/Adobe SVG, see for example here ), the directionality of the engraver's strokes also conveys useful information that might otherwise be lost from a vectorization of the photograph.

The point is perhaps even more pronounced as we move into an area of the composition where the engraver's lines do not all run horizontally:

area around mouth - photo area around mouth - engraving
Detail of mouth (photo) Detail of mouth (engraving) - 2 bits/pixel

Posterization of photo - 2 bits/pixel

Observer how detail which might ambiguously belong either to the skin or to the mustache from the perspective of the greyscale values, is unambiguously declared as one or the other by the human engraver.

Perhaps to make the point more convincingly, let's look at one of the master Dürer's own engravings [5].  It shows a fluidity of line not always found in the 19th century engravings that were done as a part of the meat and potatoes of the printing industry (see Krey's above and Morse's below):

Durer's engraving [5] of Frederick of Saxony [6]
(reduced in size -- click to enlarge)
Detail of engraving showing the engraver's lines.

Note how the contour lines convey not only shading but also, texture and depth as conveyed through contour lines running tangent to 3-dimensional curvature.

Again, back to an earlier question: so what does this mean for the modern era?

Well, vectorization of images is important. Why? Well suppose you have a thousand tanks or helicopters in a battle zone... Or for a less militaristic application, a thousand observers of a large collection of pictures, or YouTube videos that we would like to have described. Each has a limited perspective on what the on-the-ground situation actually is. We'd like to have each of those units send an annotated video-stream back to the command post via satellite or other means, so that a complete and accurate mosaic may be constructed. However, 1000 streams of live video is too rich a signal, to transmit and receive. Suppose, instead we may do local, on-the-fly processing of the images and send only vectorizations back to the command post. Those will occupy considerably less bandwidth, if the vectorizations can be completed both quickly and accurately in the local units. The problem with auto-tracing, is that ambiguous regions of pixels always interfere with scene-parsing in the semantic sense. Foreground and background often share similar color values with ill-defined perceptual boundaries. Cheeks and mustaches become indistinguishable. Houses and enemy tanks become indistinguishable.

The common approach for scene-parsing using human-computer cooperation asks our human informant (the artist/soldier on the ground -- the one possessing a visual cortex) to draw boundaries around semantically connected pixels, Suppose, instead, we instead have them draw a few simple vectors atop the bitmaps: vectors which identify the directionality of a collection of strokes that an engraver would make. These strokes would identify shared semantics, as well as the cues of depth and curvature that an engraving conveys. Maybe that is an alternative and sensible interface for quickly conveying the semantics of images. It's definitely very SVG-sounding. And from those vectors, we might not only abstract semantic information, but, in the case of a master-engraver, some 3-dimensional (or should we say 2.5 dimensional) information as well -- information that is not, strictly speaking contained in an auto-traced perspective that affiliated pixels on the basis of their chromatic similarities.

At least it worked for Albrecht Dürer.


Composite[7] of Wikimedia and Krey versions.


[1] Dürer's original painting which hangs in Munich is considered to be a very influential work in Northern European painting, discussed by a majority of the texts on that subject. The engraving here is taken from A History of Art for Classes, Art-Students, and Tourists in Europe.  8th ed. New York, A.S. Barnes & Co., 1888, 1889, 1896.  Author,  William Henry Goodyear, M.A. I scanned it at 300 dpi (greyscale), since at 200 dpi the engraving lines were not quite visible. The file was saved in GIF format, since 8 bits per pixel is about as much greyscale as web formats can hold, I think.

[2] William Patry writes (http://williampatry.blogspot.com/2005/09/albrecht-drer-and-copyright.html) "Albrecht Dürer may lay claim to the most aggressive copyright notice ever used:

Hold! You crafty ones, strangers to work, and pilferers of other men’s brains. Think not rashly to lay your thievish hands upon my works. Beware! Know you not that I have a grant from the most glorious Emperor Maximillian, that not one throughout the imperial dominion shall be allowed to print or sell fictitious imitations of these engravings? Listen! And bear in mind that if you do so, through spite or through covetousness, not only will your goods be confiscated, but your bodies also placed in mortal danger.

[3]  Compare the 8-bit / pixel scan of the original (3.2 MB) and the 2-bit / pixel scan (352 KB). Since the original engraving is in fact, 1 bit per pixel, the extra bit per pixel is responsive to misalignments between the engraver's line frequency and the scan frequency and is, in fact, useful. 8 bits instead of 2, does not appreciably increase the quality, and may, in fact, exaggerate imperfections due to aging and other paper-related printing artifacts.

[4] From http://upload.wikimedia.org/wikipedia/commons/8/8c/Albrecht_Dürer_104.jpg, as it came from the Yorck Project.

[5] Scanned (2008) from A History of Art for Classes, Art-Students, and Tourists in Europe.  8th ed. New York, A.S. Barnes & Co., 1888, 1889, 1896.  Author,  William Henry Goodyear, M.A. Scanned at 300 dpi, greyscale.

[6] According to Strauss (see below),  Dürer wrote of his subject Frederick of Saxony:

sacred to Christ. He favored the word of God with great piety, worthy to be revered by posterity forever. Albrecht Dürer made this for Duke Frederick of Saxony, Arch-Marshal, Elector of the Holy Roman Empire.

(from The Complete Engravings, Etchings and Drypoints of Albrecht Dürer By Albrecht Dürer, Walter L. Strauss Published by Courier Dover Publications, 1972). It is not immediately  clear if Dürer 's respect for the man was due to the fact that he allowed Dürer's friend Martin Luther to live, or that Frederick was apparently involved in an Imperial Privilege extended to Dürer in 1511 by Emporer Maximillian protecting "Dürer's seminal commercial response, centring on his "AD" monogram, engraving techniques that were hard to copy, privileges and law suits brought against copiers and reprinters." See Primary Sources on Copyright at http://www.copyrighthistory.org/cgi-bin/kleioc/0010/exec/ausgabe/%22d_1511b%22 . It is speculated (http://www.volcano.net/~varnbuhler/tvluth.html) that Durer's portrait of Ulrich Varnbuler may have been, in part, due that that individual's supposed support for Martin Luther, though such is not speculation typically sanctioned by the adademy.

[7] Other derivative works based on Durer may be seen in the following:

CastleRock (I need to find the proper name and source for this)
CastleRock3 (ditto above)
Bagpiper (ditto above)


Afterthought:

detail of engraving of Noah Webster by Samuel Morse, from Webster's 1909 unabridged Blurring Morse's Webster via Photoshop filters

Noah Webster, semanticist and intellectual property
theorist, engraved by Samuel F.B. Morse,
artist and inventor.[source]

Blurring Morse's Webster
via Photoshop filters

Another interesting juxtaposition of the inventor and engraver (Morse), and the copyright advocate (Webster). For Webster's involvement in the passage of the US copyright law see http://www.earlyamerica.com/earlyamerica/firsts/copyright/ .

In doing some research on Dürer, it was interesting to find how many sources on the web actually claim copyright on scans they have made of his work.  Not only does this display a blatant ignorance of SDNY's decision in Bridgman vs Corel, and therefore possibly run afoul of US Code Title 17 section 503c (which prohibits false claims of copyright), but I suspect Dürer might have demanded their heads for it when they place a defacing watermark that says "copyrighted" square over a scan of his work, without even so much as an attribution of the photographer. Perhaps I should make a "wall of shame" for the heirs to Dürer's "droits morales" so they might be able to find such "pilferers of men's brains" and, in their outrage, retaliate against that "covetousness."