Vol. IX, No. 4

February, 1997

There have been several articles in the CSA Newsletter about drawing pottery profiles. We have generally been concerned with taking existing hand drawings and finding the most efficient and accurate method for converting them into computer drawings. In most cases, this has involved scanning the hand drawings and then converting the scanned image into a CAD drawing, but we have also traced drawings directly on a digitizer. We have sought CAD drawings as the final results because of their ability to be scaled and output efficiently in a variety of formats.

The most recent article about this process appeared in the May, 1996 issue of the Newsletter ("Computerizing Pottery Profiles - Yet Again," by Alexis Menten). Ms. Menten described some of the procedures she used in experiments for her Bryn Mawr undergraduate thesis. By the time Ms. Menten had completed her thesis, she believed that the automation gained by using Draftsman® provided distinct advantages for making computer profiles from existing drawings. Using Draftsman to translate a scanned image into CAD format made it unnecessary for the computer operator to make too many choices, and she viewed that as a distinct advantage, finding the system "much easier, less stressful, and less demanding on me as illustrator." (paragraph 11) I read Ms. Menten's conclusions with interest, because my view was the reverse of her preference, I took it to be an advantage to have choices in the hands of the operator, not the computer. Therefore, I was happy to see the new version of AutoCAD®, Release 13, with its better curve-drawing process (the ability to draw truly continuous curves that are constrained by any number of points along their paths - curves called NURBS, for non-uniform rational b-spline). I thought these curves could make it possible to use AutoCAD either to draw profiles directly or, more to the point here, to trace them from scanned images of hand-drawn profiles. In either case, the NURBS curves can more accurately represent the shapes of pottery than the curves previously available either from AutoCAD or from Draftsman. Indeed, I think using the new version of AutoCAD to trace scanned drawings provides the best way to convert a hand-drawn profile to an AutoCAD profile - if the computer operator wants to make the decisions. Draftsman remains an effective system if the operator wants to make few choices. It may also enhance uniformity, since many choices are programmed into the system and would not change from operator to operator.

The new curve-drawing process may also make it easier for an artist to draw directly in AutoCAD, but my concern is with using it to copy a hand-drawn profile.

As when working with the other processes, we began with a hand-drawn profile. That was scanned, creating a computer file in one of the common drawing formats. (AutoCAD will accept scanned images in three common formats: the CompuServe format called GIF, the widely-used standard called TIFF, and the older PC format called PCX.) I used the drawing of the profile we have used in these tests all along, and I made a computer file in each of the formats acceptable to AutoCAD. Each could be imported and scaled, although determining the correct scale turned out to be more difficult than it should have been. (1)

Once imported into AutoCAD, the scaled drawing was traced two or three times for practice and then once for the profile shown here (Fig. 1)

Elapsed time was recorded. The profile was adjusted to correct some minor errors, and the time required for that was also recorded. (The time required was on a par with times recorded previously; this method should not speed or slow the process appreciably. It does reduce the cost by eliminating the need for the Draftsman program.)

In the process, I learned that the color of the scanned image mattered. I started with a black scanned line and realized that it was hard to see the traced line, even if I used another color. A yellow scanned line was more clear, but the yellow seemed unstable on screen, at least to my eyes. Finally, I tried cyan, and that worked very well. It was clear and stable, and the line drawn within the dots of the scanned image was sharp and clear.

The process requires the draftsman to draw a fine line within the rather gross one created by the scanning process. The original pencil line was relatively thick and somewhat uneven, though it seems a normal pencil line to the naked eye, so the scanned version is neither neat nor a fine line. A close-up of the scanned image (light gray dotted lines) and the line I drew as I traced (fine black line) is in Fig. 2.

It shows - assuming the printing process reproduces the image properly - that the traced line could be placed exactly in the middle of the scanned line.

One of the benefits of using this process is the positive sense of control one has. Editing the tracing shows the user exactly where the line will go and provides, as a result, excellent feedback to the user.

One problem with using AutoCAD in this way is the file size that results. With only one scanned image and the tracing of it, the file was nearly 1.5 MB in size; but even three versions of the tracing without the scanned image did not require a half MB of disk space. (In fact, adding additional profiles to the file only added about 10 KB each.) Therefore, one would not wish to keep many scanned images in a file for long. Of course, once the scanned image has been traced, it need not be kept in the file.

The resulting drawing, shown here in Fig. 3, is on a par with the others we have created. I doubt that the quality difference is large. There are two important differences, though. First, the profile is a true, continuous curve, not a series of connected curves. Second, as stated above, I felt that the adjustment process was superior; so it seemed to me that the results were as well. I will confess, however, that the difference may not show at a scale smaller than one-to-one.

For other Newsletter articles concerning pottery profiles and capacity calculations, consult the subject index.

Next Article: Addition to Archaeological Data Archives

Table of Contents for the Feb, 1997 issue of the CSA Newsletter (Vol. 9, no. 4)

Table of Contents for all CSA Newsletter issues on the Web