Clipped contrast

Clipped contrast, a part of algorithm transparency, is computed during feature extraction for every primary block on the block map by taking smoothed histogram for that block, clipping some fraction of brightest and darkest pixels, and taking the difference between brightest and darkest pixel in the remaining histogram. Clipped contrast is used to compute absolute contrast mask and relative contrast mask.



Contrast visualization can be computed from this stage's binary data combined with block map and original fingerprint image. Visualization itself is not part of transparency data.

Contrast visualized as green circles, one per block, with fingerprint image in the background
Contrast map. Circle's surface area is proportional to contrast within that block.

Binary data

Binary data is a 2D array with dimensions that can be found in the associated json data, height first, and that are also equal to the number of primary blocks in the block map.

Binary data consists of rows of primary blocks ordered from top to bottom. Every row is a sequence of blocks from left to right. Every block is a 64-bit IEEE754-encoded big-endian floating-point number in range from 0 (no contrast) to 1 (maximum contrast).

Example: 011-clipped-contrast.dat (650 scalars, range 0 to 0.99, average 0.33)


Json data describes basic properties of the binary data, most importantly dimensions of the contrast map, which is the same as the number of primary blocks.

Example: 010-clipped-contrast.json

  "axes": [
  "dimensions": [
  "scalar": "double",
  "bitness": 64,
  "endianness": "big",
  "format": "IEEE754"