Thanks to my former colleague Helge Holzmann, who works at the Internet Archive, we are now able to analyse the CIS/COD images that have been archived by the Internet Archive. Helge extracted a list of 1552 URLs that hosted files with the MIME type “image/cis-cod” from the Internet Archive’s collection.

This list is based on the MIME type which the HTTP servers delivered in their HTTP response’s “Content-Type” field during crawling. So I was curious as to how many of those files are actually CIS/COD images. The first step in finding that out is to download these files and this worked for 1230 of the 1552 listed files.

The script actually does a bit more, by already checking the first three bytes of each file for the magic marker “CIS”:

three=$(head -c3 "${tmpfname}")
if [ "${three}" = "CIS" ]; then
    mimetype="image_cis-cod"
fi

It then assigns the MIME type “image/cis-cod” to those files and for all other files uses file --brief --mime-type to determine the MIME type. The result is as follows:

So only 221 of the 1552 candidate files could be downloaded and are actually (or at least very likely) CIS/COD images.

In which version of the file format have the images been stored? Looking at their first six bytes reveals that:

for f in image_cis-cod/*cod; do
    echo $(head --bytes=6 --quiet $f)
done | sort | uniq -c

So these are mainly files in CIS/COD version 2.5. With a magic file for the file command we can also extract the heights and widths of the images and plot them:

So most images are smaller than 800x600 pixels and there is one large image 2000x1400 pixels in size.

Let us have a look at the hosts that hosted the images

Apart from three hosts related to companies that developed the CIS/COD image format (compsci.com, infinop.com, euclid.infinop.com) two British web pages have hosted a larger amount of CIS/COD images: british-waterways.com and highpeaknet.com.

Unfortunately, there is no tool (yet) for converting the images into a useable image format. So I converted the 42 images in CIS/COD version 3.0 manually with LSIC 2.6 and created the following collage:

These might be the only CIS/COD 3.0 images that are available on the Web!

Infinop, Inc. and Computer and Information Sciences, Inc.

The Lightning Strike image codec was distributed by a company named Infinop, Inc. and there are several links between that company and Computer and Information Sciences, Inc. I first stumbled over the latter name (that can be abbreviated as CIS, Inc.) in the publication Dissemination of compressed Satellite imagery within the Navy SPAWAR Central Site Product Display environment (archived tar.gz with all HTML documents from the host www.compsci.com) that was published on October 1, 1995, at The 1995 Science Information Management and Data Compression Workshop. Looking at the company’s archived web page we see a directory listing with two folders – infinop and lstrike that link CIS to Infinop and the Lightning Strike image codec. The windows subdirectory of the “infinop” directory contains the executables ls16tiny.exe, ls32tiny.exe, and lstplug.exe that contain 16 and 32 bit variants of the Lightning Strike web browser plugin in version 1.7, the file lsic25.zip that contains the Lightning Strike Image Compressor (LSIC) in version 2.5, as well as a directory with several exemplary CIS/COD images.

Another evidence for the relationship between the two companies are their addresses. The Infinop web page states the company’s address as 3401 East University #104, Denton, TX 76208. The above-mentioned publication states as the address of Compsci, Inc. 303 N. Carroll, Suite 108, Denton TX 76201. Both locations are roughly 5 miles apart in Denton, Texas..

Another link between the two companies is via the former homepage of Hongyang Chao, which was hosted on compsci.com and states that she is the designer of the Lightning Strike image compressor, linking to infinop.com.

Looking at other people, a press release states that “Craig Fisher” is “president of Infinop, Inc.” and the authors of the above-mentioned publication are “Oleg Kiselyov and Paul Fisher”, who also co-authored other image compression-related papers, for example, Image Compression with Iterated Function Systems, Finite Automata and Zerotrees: Grand Unification (submitted to arXiv on March 15, 2000). Oleg Kiselyov graduated as a PhD under supervision of Dr. Paul S. Fisher at the university of North Texas. We could speculate whether Paul (S.) Fisher and Craig Fisher might have been relatives.

Infinitron Research International, Inc.

There’s another company – Infinitron Research International, Inc. – that seems to have played a role in the codec. It is mentioned in the patent WO1998040842A1 that was published on September 17, 1998. The patent contains the paper An approach of fast integer reversible wavelet transform for image compression written by Hongyang Chao and Paul Fisher of “Computer and Information Science Inc.” [sic] located at 3401 East University #104, Denton, TX 76208. That is the address that Infinop, Inc. (!) gave on its web page and not the CSI, Inc. page from the other paper! The patent also contains a product flyer for the Lightning Strike Image Compressor (LSIC) version 3.0, distributed by Infinitron, Inc. – also located at 3401 East University #104, Denton, TX 76208. And, last but not least, the patent shows HTML source code to embed an image map that links to www.infinop.com.

Overview and Summary

Given the many clues it seems evident that Infinop, Inc., Infinitron Research International, Inc., and Computer and Information Sciences, Inc. were all related to the Lightning Strike image compression technology and were likely founded or run by the same (group of) people.

About the name

According to archived web pages and the ArchiveTeam wiki the image format is called “Lightning Strike”. The wiki states the MIME type to be “image/cis-cod”. The three letters “CIS” are the first three bytes of each image file (a magic number) and at the moment it is not clear, what they stand for. The files typically have the file extension “cod” which very likely stands for “coded” (image). For the purpose of this web site I have decided to use the more technical term “CIS/COD” which is derived from the MIME type and refers to the two main indicators of the image files.

Further reading

At the moment the best overview is given by the ArchiveTeam wiki and a good starting point for further reading are the following blog posts:

• Finding the tiles – where we first discover that CIS/COD is used to store satellite images in D-Sat 1.
• Learning about the image format – which provides first results of some (mainly Internet Archive) research on finding resources describing the image format.
• Viewing tiles with Firefox – which is a practical guide on how to view (and convert) images with an old Windows version of Firefox.
• Fragmenting the tile data – where I tried to understand the arrangement of the data for the wavelet sub-bands.

On this site, the documentation page currently mainly lists information about the known software and DLLs and the resources page provides access to a magic file for the file command and a file format description for Kaitai Struct.