v1.2.0

open-iris v1.2.0

Changelog

Major

Added

• SharpnessEstimation in eye property estimation and in the pipeline. SharpnessValidator uses threshold min_sharpness to reject very blurry images.
• FragileBitRefinement in the pipeline. This will result in better templates by removing noise. But may also increase validator error fromIsMaskTooSmallValidator as more bits are masked out in the template.

Changed

• rewriteFragileBitRefinement that includes 1) added thresholding based on angle in polar method. 2) added thresholding in cartesian method. 3) separate masking in the imaginary and real part of the filter response. 4) normalization of filter response (to handle non-normalised filters such as fixed-point filters).
• convert mask_response to complex array due to separate masking in the imaginary and real part.
• Refactor parameters for HammingDistanceMatcher, SimpleHammingDistanceMatcher and the underlying utils function hamming_distance. Remove parameter norm_nb_bits in SimpleHammingDistanceMatcher and nm_type in HammingDistanceMatcher. Add parameter norm_gradient to SimpleHammingDistanceMatcher and norm_gradient and separate_half_matching to HammingDistanceMatcher. Change the normalized Hamming Distance estimation by replacing the square-root term by linear approximation. These changes result in simpler calculation and slightly better performance.

Minor

Added

• detailed error types to Pupil2IrisPropertyValidator, namely Pupil2IrisValidatorError_Constriction, Pupil2IrisValidatorError_Dilation, Pupil2IrisValidatorError_Offcenter.

Changed

• changed condo environment dependency on python from version 3.8 to 3.9