This paper presents a pupil detection algorithm based on a global adaptive thresholding procedure. The binarization threshold for each image is determined by analyzing the derivative of the percentile function. Additional morphological transformations are used to obtain the set of possible pupil contours. Afterwards, the algorithm selects the largest compact contour, fits an ellipse to it by using the least square method and delivers its center as the output. By using this method, the algorithm achieves good results even in non-uniform and variable lighting conditions. For two representative databases of eye images, the 5 pixels detection rate is higher than 92%.

In this paper, the implementation of a new pupil detection system based on artificial intelligence techniques suitable for real-time and real-word applications is presented. The proposed AI-based pupil detection system uses a classifier implemented with slim-type neural networks, with its classes being defined according to the possible positions of the pupil within the eye image. In order to reduce the complexity of the neural network, a new parallel architecture is used in which two independent classifiers deliver the pupil center coordinates. The training, testing, and validation of the proposed system were performed using almost 40,000 eye images with a resolution of 320 × 240 pixels and coming from 20 different databases, with a high degree of generality. The experimental results show a detection rate of 96.29% at five pixels with a standard deviation of 3.38 pixels for all eye images from all databases and a processing speed of 100 frames/s. These results indicate both high accuracy and high processing speed, and they allow us to use the proposed solution for different real-time applications in variable and non-uniform lighting conditions, in fields such as assistive technology to communicate with neuromotor-disabled patients by using eye typing, in computer gaming, and in the automotive industry for increasing traffic safety by monitoring the driver’s cognitive state.

A structure that has a weak-inversion biased MOS transistor at its core which can be used to build analog multipliers, dividers, squarers or square rooters is presented. It relies on the exponential dependence of the drain current to the gate-to-source voltage of a MOS transistor that is biased in the weak-inversion region. The proposed circuit produces at the output a current proportional to the gate-to-source voltage that has, conversely, a logarithmic dependence on the drain current, which is set to be the input current. Given that linear operations in the log-domain correspond to exponentiations, multiplications or divisions in the antilog-domain, the proposed structure serves as a building block for analog multipliers, dividers and many other non-linear circuits.

This paper presents a CMOS circuit able to magnify the value of a reference bias current. The magnification factor can be programmed via a 10bit interface in the 1-1024 range. A time domain technique is used to compare a replica of the output current and a replica of the reference current. A negative feedback circuit minimizes the errors up to 1%.

In this paper a new pupil detection algorithm based on circular Hough transform approaches is presented. The proposed algorithm includes an adaptive quantitative binarization procedure, a pupil reconstruction stage by using certain morphological operations and the image contour detection performed with the Laplace operator. To save the computing resources, the circular Hough transform is applied on the edge of the binarized eye image and provides better results for noisy eye image captured in nonuniform and variable lighting conditions. The CHT based algorithm achieves a detection rate at five pixels higher than 82% for three different and highly representative databases, totalizing 1324 eye images.

In this paper a new pupil detection algorithm appropriate for development of head mounted eye tracking interfaces is presented. The robustness of the proposed algorithm is given by its low running time and the ability to accurately provide the coordinates of the pupil center even under real-world scenarios like variable illumination and high noise conditions. The proposed algorithm is based on an adaptive quantitative binarization technique with a feedback-loop controlled threshold. The pupil center is detected by applying the Least Square Fit to Ellipse procedure to the processed eye image. The performance of this algorithm is given by a 5 pixels detection rate of 93%, obtained for the eye images from two representative databases.