Pixelwise orientation

Pixelwise orientation, a part of algorithm transparency, contains orientation vector for every pixel of the image. Orientation vector encodes orientation angle (essentially an inclination angle of straight line) along with the strength of orientation signal in that pixel. Pixelwise orientation is computed from equalized image during feature extraction. It is processed into block orientation.

KeywordSuffixZip
pixelwise-orientation.json022-pixelwise-orientation.json
pixelwise-orientation.dat023-pixelwise-orientation.dat

Visualization

Visualization of pixelwise orientation was constructed from this stage's binary data alone. The apparent fingerprint structure is implicitly present in the data due to the nature of orientation detector rather than being explicitly added to the visualization. Visualization itself is not part of transparency data.

Pixelwise ridge orientations on a fingerprint visualized as hue from horizontal red to vertical green
Pixelwise ridge orientation. Horizontal orientation is mapped to red color, vertical to cyan. Strength of orientation signal is mapped to color saturation. Pixels with no orientation data are white.

Binary data

Image size can be obtained from dimensions field in the associated json data, height first, or from pixels field in block map. Binary data consists of rows of pixels ordered from top to bottom. Every row consists of pixels ordered left to right. Every pixel is a vector of two 64-bit IEEE754-encoded big-endian floating-point numbers (X and Y axes of the vector).

Vectors represent special orientation angles defined below. Length of the vector indicates strength or quality of angle data for that pixel. Vectors have they Y axis pointing down and X axis pointing right. Vector can be converted to orientation angle in range from 0 to 360 degrees by assuming that vector for zero angle points right and angle vectors rotate clockwise with increasing angle. Areas outside the mask or with no orientation information are zeroed.

Orientation angles are something completely different from direction angles. For example, right and left directions have the same horizontal orientation. Orientation can be thought of as an infinite straight line. Zero orientation angle represents horizontal orientation and increasing the angle turns the orientation clockwise. Orientation angle of 180 degrees represents vertical orientation. Orientation then increases further clockwise and reaches horizontal orientation again at 360 degrees.

Example: 023-pixelwise-orientation.dat (290,224 scalars, range -11.8 to 14, average -0.22)

Json

Json data describes basic properties of the binary data, most importantly dimensions of the image.

Example: 022-pixelwise-orientation.json

{
  "axes": [
    "y",
    "x",
    "axis"
  ],
  "dimensions": [
    374,
    388,
    2
  ],
  "scalar": "double",
  "bitness": 64,
  "endianness": "big",
  "format": "IEEE754"
}