Understanding the subject-specific effects of pupil dilation on iris recognition in the NIR spectrum

Abstract

Abstract-A recent report from the National Institute of Standards and Technology (NIST) showed that changes in pupil dilation affect the performance of iris recognition algorithms. Hence, there is a need to explore the effects of pupil motion from a biological standpoint. Our work looks at the pupil's response to light, otherwise known as the pupil light reflex (PLR). By modeling the PLR using a nonlinear delay differential equation while considering images acquired in the near infrared (NIR) spectral band, we study both average and subject-specific pupil dilation effects. Experiments conducted on the WVU iris video dataset 1 convey the efficacy of our work in describing and evaluating pupillary response for both general and individual responses. The results of this work can be used to develop robust iris recognition algorithms that handle the effects of pupil dilation.

The states of dilation in iris recognition (a preliminary study)

“…Their results showed that, when image comparisons are made between the extreme degrees of dilation 1 to the average dilation, bit errors increased by over 10%. Furthermore, from a dilation viewpoint, Gejji et al [11] revisited the work of [12] to abstractly express the effects of dilation in the visible spectrum and the NIR spectrum. Their work showed that, for the case of steady illumination, Pamplona's model to represent the typical pupillary response, can be expressed as…”

“…Our motivation for this work comes from the literature in [4], [9], [10] and [11] that suggest a need to develop an analytical and universal approach to describe the effects of dilation on iris recognition caused by ambient lighting. The works of [4] and [10] explore this concept from a template perspective; however, this is mainly from an empirical point of view, which is dependent on the technology and data at hand and does not fully consider the physiological effects of the pupil.…”

“…The works of [4] and [10] explore this concept from a template perspective; however, this is mainly from an empirical point of view, which is dependent on the technology and data at hand and does not fully consider the physiological effects of the pupil. Additionally, the development of the mathematical models in [9] and [11] illustrate that the iris muscle activity, governed by the sphincter and dilator muscles, determines changes in both pupil size and iris deformation. These changes, described in the works of [4], [10] and [14], indicate that they also affect the associated iris templates.…”

“…First, we build our theoretical model by considering the empiric formulation of [4] to explore the relationship between the state of the input iris image (hitherto referred to as the acquired state) to the various stored states in an iris recognition database for an average subject. We do this by incorporating the properties of transition processes with the pupillary response, described in [15], [16] and [11], where the result is a transport equation with random initial conditions. Second, we perform theoretical analyses of our transport equation where we consider evolutionary solutions and an approximate solution to our model by considering short temporal duration.…”

NIRExpNet: Three-Stream 3D Convolutional Neural Network for Near Infrared Facial Expression Recognition

“…Gejji, R.S. et al [13] modeled the pupil's response to light by using a nonlinear delay differential equation in the NIR spectral band, which can be used to develop robust iris recognition algorithms. Son, C. et al [14] captured NIR images and visible color images and presented a new NIR dehazing model based on color regularization, which can solve the color distortion and haze degradation problems at the same time.…”