ImageBio2

  1. What do we observe in irises that tends to be unique between human?
    • Unikatna struktura mišičnih vlaken, ožilja in živcov
    • Image Upload 1
  2. Who is to credit for the initial breaktrough of iris recognition?
    • John Daugman
    • Image Upload 2
  3. Why do we use the near-IR spectrum for iris recognition?
    Near-IR illumination omogoča zajem uporabnih slik tudi, ki imamo dark irises(Ker pri navadni svetlobi bi se tekstura zlila z pupil, kar bi naredilo segmentacijo težjo)

    • Image Upload 3
    • Kot vidimo barva iris sploh nima nekega upliva na zajem pri near-IR spectrum.
  4. What is the pupillar boundary?
    • Boundary between the pupil and the iris.
    • Image Upload 4
  5. What is the limbus boundary?
    • Boundary between the iris and the sclea.
    • To je zunanje območje od iris.
    • Image Upload 5
  6. What is noise mask (in iris recognition)
    • Maska, ki označi neželene lokacije, kot so:
    • trepalnice
    • sence
    • odsevi
    • Image Upload 6
  7. What is the Daughman's rubber sheet model?
    • A form of iris normalisation in which we unfurl (razvijemo) circular iris into a standard rectangular image (katero kasneje uporabimo za encoding)
    • Image Upload 7
  8. What is the result of applying Gabor filter on the iris image?
    Location of the iris image points into a complex plane.
  9. How do we encode complex values to obtain the iris code?
    • Pogledamo v katerem kvadrantu se nahaja, lahko gledamo tudi kot od x osi.
    • Image Upload 8
  10. How big is iris code typically?
    • Standard commercial iris code is 2048 bits (plus 2048-bit mask)
    • 2048/2 = 1024 sample points in the iris.

    1024 = 8 radii * 128 samples

    Ta rubbersheet model, ko ga raztegnemo je velik 128 x 8

    Image Upload 9
  11. Why do we use Gabor filters in iris encoding?
    The response of the Gabor filter seems to mimic human vision (unproven)
  12. Which measures do we use for matching iris encodings?
    • Hamming distance (XOR operator).
    • Grem po vrsti in preverjam z XOR in pišem enke kjer so razlike.
    • Image Upload 10
  13. Which operator on the iris code corresponds to rotation of the iris on the image.
    Ker se glava premika in je lahko postrani, izračunamo metriko, ki se imenuje circular shift.

    To naredimo tako, da računamo vse shifte in obdržimo najmanjšega.


    Problem je, ker to poveča auntentikacijski čas, ker računamo več rotacij.

    Popravimo autentikacijsko distribucijo in poslabšamo impostor distribution.
  14. What can we observe in the Fourier spectrum if our image contains out of focus blur?
    • Sestavljeno je predvsem iz nizkih frekvenc.
    • Image Upload 11
  15. What can we observe in the Fourier spectrum  if our image contains motion blur?
    • Primankuje srednjih in visokih frekvenc, zelo podobno kot out of focus.
    • Image Upload 12
  16. What can we observe in the Fourier spectrum if our image contains a lot of occlusions?
    • Vsebuje veliko sredjih in visokih frekvenc.
    • Image Upload 13
  17. Does the pupil dialation/contraction affect the quality of authentication?
    Da, razlika pri primerjanju raztegnjene in skrčene je velika in poveča FNMR (False Non Match Rate).
  18. Does lenses affect iris authentication?
    • It depends on lens manufacturer but largely yes.
    • Lahko pa shranimo oblike leč v training set, katere potem sistem prepozna.
    • Fake leče pa se da tudi zaznati če izračunamo fourier spectrum.
  19. How may we detect contact lenses using the fourier spectrum?
    • Natisnjen iris vsebuje neke tipične artifakte, katere se da zaznati z 2D fourier spectrum.
    • Image Upload 14
  20. How robust are irises to ageing?
    There is degradation, but single enrollment can last you a lifetime.
Author
wolf
ID
354408
Card Set
ImageBio2
Description
fdsafsdafsad
Updated