This WSQ testing is updated for "Wavelet Scalar Quantization (WSQ) Gray-Scale Fingerprint Image Compression Specification Version 3.1" which is available at https://www.fbibiospecs.cjis.gov/ . The WSQ algorithm was developed by the FBI in close coordination with NIST in the early 90's.
The reference image dataset can be downloaded from NIST at either of the following:
The difference between them is that the former contains uncompressed fingerprint images in a headerless raster format, while the latter contains those images in Portable Graymap (PGM) format, which is documented at:
To use the raw-format version, the image dimensions are listed in the Raw Image Dimensions section below.
Version 2.0 of the reference image dataset is a superset of the previously-available dataset. It has been expanded to include slap, live-scan, and low- and medium-quality rolled fingerprint images.
The certification process, described in more detail below, entails encoding the unencoded images in the dataset, decoding the encoded images in it, and returning the output files to NIST for analysis.
Once the vendor has determined that their software meets all the requirements in the WSQ Specification (v3.1), one or more FBI WSQ software implementation numbers should be obtained from:
Stan Janet <sjanet [at] nist.gov (subject: WSQ) (sjanet[at]nist[dot]gov)>
In the email, specify the platforms that certification is being sought for. A unique software implementation number will be assigned to each platform. A platform is taken to be the computer hardware (computer make & model, and CPU make & model), operating system, compiler (name & version), and any virtual machine(name, version and configuration), if used, that the output images were generated with.
The 32-bit and 64-bit versions of an operating system are considered different platforms.
Different Service Packs of the same Microsoft Windows version have not been considered different platforms. However, please specify the Service Pack level used.
The encoder test set contains 40 images, each of which must be compressed at two different bit rates, producing 80 separate tests for the encoder. For each test, the encoder under test must match the file size, the parameter values in the header, and the quantized wavelet coefficient bin index values of the reference data within the accuracy requirements contained in the WSQ Specification (v3.1). The WSQ encoder algorithm (parameterization) must be set in the Frame Header by setting the "Ev" field to "2" (FBI Encoder Number Two) as described on page 20 of the specification (v3.1). The software implementation number must be placed in the "Sf" field of the Frame Header. The WSQ encoder algorithm and software implementation number must be contained in the Frame Header of each WSQ-format file.
The decoder test set contains 2 versions of the same 40 images as the encoder test set, but in WSQ-format, having been compressed by NIST at the same 2 bit rates, as well as 6 more images compressed with filter values other then those given in Part III of the WSQ Specification (v3.1). For each of these86 tests, the reference-compressed image is reconstructed using the decoder under test and the result compared to the reference reconstruction. The 6 *.wsq files in the "decode/not_7_9" directory were compressed using filter tap sets of lengths other than 7/9 at bit allocation target of 0.75 (compression ratio 15:1).
For each platform:
Then, for all platforms, the output files go to NIST for analysis. The fastest and easiest way is to put them in a tar (or zip) archive placed on a website and email its URL to Stan Janet. Alternatively, they can be burned to CD-R/DVD-R/DVD+R and shipped to:
Stan Janet - WSQ Certification Liaison
100 Bureau Drive, Mailstop 8940
Gaithersburg, MD 20899 USA
If multiple submissions are in the same archive or on the same disc, make sure they are in separate top-level directories with each directory name containing the corresponding software implementation number.
The *.pgm output files should be in the smaller, binary "P5" form instead of the larger, ASCII "P2" form.
Do not return the original *.raw and *.wsq files from the reference image dataset.
A text file listing the vendor contact information (email addresses of those to contact if there are issues with the submission, as well as the full mailing address of who the FBI should send a letter confirming eventual certification), the software implementation number, the computer platform information (hardware, operating system, and compiler), and the vendor's internal version identifier for the implementation (if there is one),should also be present in the archive or on the disc. The files should contain ASCII-7 characters only.
After processing the appropriate files for the encoder and decoder, a certification request containing the test report and test results, and all generated compressed and reconstructed files (cmp000xx with extensions '.wsq' and '.pgm') are forwarded to the FBI for review and evaluation.
The FBI may request additional information or conduct supplemental tests to determine full compliance with the WSQ Specification (v3.1).
If all the required information is complete and the results are satisfactory, the FBI CJIS Division will issue a letter certifying that the implementation is compliant with the WSQ Specification (v3.1). That implementation number and description will than be added to the current list of approved implementations maintained by the FBI.
The FBI certification for an implementation will apply only to a specific configuration. A configuration encompasses the software version of the encoder/decoder, hardware platform, operating system, and compiler used. As any of these components change, a recertification will be required.
Raw Image Dimensions
Filename Width Height
cmp00001.raw 589 605
cmp00002.raw 832 768
cmp00003.raw 832 768
cmp00004.raw 815 752
cmp00005.raw 832 768
cmp00006.raw 832 768
cmp00007.raw 545 638
cmp00008.raw 800 750
cmp00009.raw 512 678
cmp00010.raw 375 526
cmp00011.raw 614 717
cmp00012.raw 516 716
cmp00013.raw 539 651
cmp00014.raw 466 578
cmp00015.raw 436 670
cmp00016.raw 666 758
cmp00017.raw 521 666
cmp00018.raw 750 750
cmp00019.raw 750 750
sample_01.raw 1600 1500
sample_03.raw 1600 1500
sample_04.raw 1600 1500
sample_10.raw 1600 1500
sample_11.raw 1600 1500
sample_19.raw 1600 1500
a001.raw 1508 1008
a002.raw 1452 1006
a018.raw 1503 1021
a039.raw 460 996
a070.raw 1580 1016
a076.raw 799 767
a089.raw 1586 993
a107.raw 1544 1206
a129.raw 398 919
a165.raw 404 958
b082.raw 416 989
b124.raw 1535 984
b157.raw 1521 1034
b158.raw 399 965
b186.raw 398 933