This page supplements an earlier set of JPEG-2000 compression tests, completed in November 2005, which also made a case study of antique maps.

For the first set of trials, the two TIFF files I compressed were lower-resolution images (one grayscale, one true color), of the sort deemed more appropriate for Web publication when file size and download time are the primary concerns. In general, I follow a guiding principle that the quality of compressed images I post to the Web should be good enough for serious research and scholarship, but of as small a file size as possible. The sort of high resolution required for peer-reviewed publication or intense scholarly scrutiny is not, in most cases, a necessity. Hence, I try to keep all enlargements within a range of 1000–3000 pixels wide, with a 1000-pixel width, the norm.

There are, however, select cases where a higher resolution than the norm is desirable, and the image tested here — 8040x5702 pixels — falls in that category. With so many pixels to describe, the starting file size is a hefty 131.18MB.

This test image of what is now known as “the Velasco map” was prepared upon request for a colleague who has need of the higher resolution. Indeed, two lower-resolution versions of the image have already been posted to she-philosopher.com —

• the 740x526 pixel image that appears at the top of the first of five Gallery Exhibits on color and/or modern reproductions of the Velasco Map, and
  
  
• the 2650x1884 pixel image accessible from the “View an enlarged 2650 x 1884 pixel JPG image” link immediately below
  

— and are, I believe, more than adequate for most purposes. By way of comparison, the latter lower-resolution 2650x1884 pixel image is a 1,039,030 byte (0.99MB) JPG file, which I compressed from a starting TIFF file of 14.30MB using Photoshop’s “Save for Web” JPG codec (my driver of choice for the bulk of images I post to the Web).

Given that the high-resolution 131.18MB TIFF file tested here is about 9 times the size of the TIFF file used to produce the lower-resolution versions of the Velasco Map already posted, I expected the JPEG-2000 compression format to substantially out-perform standard JPEG, GIF and PNG formats.

It did not.

Indeed, the default settings in the LEAD Technologies plug-in for Photoshop produced a compressed file of 5.25MB, slightly larger than the 4.85MB file I was able to produce with Photoshop’s “Save for Web” JPG codec.

As before, optimal compression requires experimentation, and manual tweaking of the JPEG-2000 driver’s standard and advanced settings. Having done this, I was able to come up with a compressed file size of 2.62MB, but even this file size is matched by Photoshop’s “Save for Web” JPG driver, using its LOW setting (2.78MB). Unfortunately, I could not ascertain the quality of this compressed file, as Photoshop was unable to write the file to disk without using more memory than my system has available. I suspect image quality would be comparable, though.

For the current set of tests, I have used knowledge gained from the first go-round to streamline the process a little (a necessity, given the large size of the files, both compressed and uncompressed).

As user feedback has shown the files produced with LEAD Technologies’ JPEG-2000 plug-in for Photoshop to be most reliable, this time around, I limited my tests to this particular codec. I also chose to restrict my JPEG-2000 file types to the standard JP2 (i.e., no J2K tests were made).

Although I created a range of test files using IrfanView’s JPEG driver, I decided to discard these as none of them met the quality of the comparably-sized JPEG files generated with Photoshop. Hence, no references to these IrfanView-generated JPEG files appear below in the table of results or data sets.

While I have included results for PNG and GIF compression schemes in the table that follows, I have not included links to any sample files. The PNG test files I generated were still way too large for posting to the Web (the smallest being a hefty 21.57MB), and all showed some loss of detail. Similarly, the one GIF test file I created was even larger (24.61MB), and had even more noticeable loss of detail (probably due to the driver’s 256-color limit).

In general, the JPEG-2000 files take longer to load than comparably-sized JPEG files (e.g., 11 seconds for a 5.25MB JP2 file versus 3.5 seconds for a 5.70MB JPG file). While an 11-second load time is long enough to require a user caveat, my 2.62MB JPEG-2000 file of choice loaded in a speedy 0.7 seconds, alleviating this concern.

In sum, I continue to question the efficacy of using the JPEG-2000 compression scheme on antique maps of the kind tested here.

It would appear that the much-touted JPEG-2000 advantage pertains mostly to more complex image structures (e.g., continuous tone photographs) than is typical of the 17th-century maps and engravings with which I tend to work.