Kang A. Choi

Eyeball model-based iris center localization for visible image-based eye-gaze tracking systems

In general, the visible image-based eye-gaze tracking system is heavily dependent on the accuracy of the iris center (IC) localization. In this paper, we propose a novel IC localization method based on the fact that the elliptical shape (ES) of the iris varies according to the rotation of the eyeball. We use the spherical model of the human eyeball and estimate the radius of the iris from the frontal and upright-view image of the eye. By projecting the eyeball rotated in pitch and yaw onto the 2-D plane, a certain number of the ESs of the iris and their corresponding IC locations are generated and registered as a database (DB). Finally, the location of IC is detected by matching the ES of the iris of the input eye image with the ES candidates in the DB. Moreover, combined with facial landmark points-based image rectification, the proposed IC localization method can successfully operate under natural head movement. Experimental results in terms of the IC localization and gaze tracking show that the proposed method achieves superior performance compared with conventional ones.

Improved pupil center localization method for eye-gaze tracking-based human-device interaction

This paper presents an improved pupil center (PC) localization method for eye-gaze tracking. In the proposed method, an input infrared eye image is repeatedly binarized with a finite number of different thresholds to produce a stack of binary images. Among all the blobs, which are groups of connected binary pixels in the binary image stack, we find a blob whose shape is the most similar to pupil, in terms of the size, the aspect ratio, the moments, and the circularity. Consequently, the centroid of the final resultant blob is regarded as the PC location. Experimental results show that the proposed method outperforms conventional ones.

Improved remote gaze estimation using corneal reflection-adaptive geometric transforms

Abstract. Recently, the remote gaze estimation (RGE) technique has been widely applied to consumer devices as a more natural interface. In general, the conventional RGE method estimates a user’s point of gaze using a geometric transform, which represents the relationship between several infrared (IR) light sources and their corresponding corneal reflections (CRs) in the eye image. Among various methods, the homography normalization (HN) method achieves state-of-the-art performance. However, the geometric transform of the HN method requiring four CRs is infeasible for the case when fewer than four CRs are available. To solve this problem, this paper proposes a new RGE method based on three alternative geometric transforms, which are adaptive to the number of CRs. Unlike the HN method, the proposed method not only can operate with two or three CRs, but can also provide superior accuracy. To further enhance the performance, an effective error correction method is also proposed. By combining the introduced transforms with the error-correction method, the proposed method not only provides high accuracy and robustness for gaze estimation, but also allows for a more flexible system setup with a different number of IR light sources. Experimental results demonstrate the effectiveness of the proposed method.

A novel iris center localization based on circle fitting using radially sampled features

In this paper, we propose a novel iris center (IC) localization method. Given an eye region image, we first find an initial iris center and perform iterative circle fitting until the center of the circle converges on the final IC. In each iteration step, a certain number of feature points (FPs) are extracted from radial sectors defined around the IC, and then grouped into several clusters for circle fitting. Through the iteration, the circle center converges on the final IC. Experimental results on the public dataset GI4E show that the proposed method achieves the superior performance as compared with conventional ones.

A novel 2-D mapping-based remote eye gaze tracking method using two IR light sources

Consumers nowadays demand for more convenient human-device interaction (HDI) in every device they use. As a part of the HDI technologies, remote eye gaze tracking (REGT) is getting more attentions. Among state-of-the-art REGT methods, the ones which utilize four infrared (IR) lights have better accuracy and robustness to head movements. In this paper, a novel glint estimation method is proposed, which enables four-glint-based REGT methods to operate with only two IR light sources. Given two glints obtained from IR light sources attached on the upper or lower corners of the screen, the proposed method estimates the other two virtual glints based on mathematical and geometric principles. Then, the point of gaze (POG) can be estimated using a mapping function established between the four glints and the screen through a user calibration process. Experimental results show that the REGT system using the proposed glint estimation method can achieve highly competitive performance compared with the ones which use four IR light sources.

A novel iris center localization method based on the spherical eyeball rotation model for human-device interaction

This paper presents a novel iris center (IC) localization algorithm for human-device interaction (HDI). In the proposed method, many different IC locations and their corresponding iris boundaries (IBs) are first registered as a database. Using the database, IB matching is performed to locate the IC. Experimental results show that the proposed algorithm outperforms the conventional ones especially when the iris is located at the corners of the eye.

Robust remote gaze estimation method based on multiple geometric transforms

Abstract. The remote gaze estimation (RGE) technique has been widely used as a natural interface in consumer electronic devices for decades. Although outstanding outcomes on RGE have been recently reported in the literature, tracking gaze under large head movements is still an unsolved problem. General RGE methods estimate a user’s point of gaze (POG) using a mapping function representing the relationship between several infrared light sources and their corresponding corneal reflections (CRs) in the eye image. However, the minimum number of available CRs required for a valid POG estimation cannot be satisfied in those methods because the CRs often tend to be distorted or disappeared inevitably under the unconstrained eye and head movements. To overcome this problem, a multiple-transform-based method is proposed. In the proposed method, through three different geometric transform-based normalization processes, several nonlinear mapping functions are simultaneously obtained in the calibration process and then used to estimate the POG. The geometric transforms and mapping functions can be alternatively employed according to the number of available CRs even under large head movement. Experimental results on six subjects demonstrate the effectiveness of the proposed method.

A novel single IR light based gaze estimation method using virtual glints

Among various infrared (IR) image based remote eye gaze tracking (REGT) methods, 2-D interpolation based methods, e.g., the cross ratio-based and homography normalization (HN)-based methods, have been widely researched owing to their highly accurate performance and simplicity with respect to hardware setup. These methods, however, can be employed only when user utilizes more than four IR light sources to obtain multiple corneal reflections projected on the image plane, i.e. glints. In this paper, a novel 2-D interpolation-based REGT method with a single IR light source is proposed, which attains both accuracy and headpose robustness. In the proposed method, the virtual glint (VG), which can substitute for the actual glint, is first estimated by utilizing the mathematical and geometrical principles established between the 3-D location of an IR light source and the corresponding glint in the image plane, and then the point of gaze (POG) is calculated by employing the HN method using the estimated VGs. The experimental results indicate that the proposed REGT method is highly competitive with conventional ones requiring multiple IR light sources, in terms of accuracy and robustness against head movements.

Improving the usability of remote eye gaze tracking for human-device interaction

This paper presents a novel single-point calibration-based remote eye gaze tracking (REGT) method. The proposed method consists primarily of two steps. First, a user calibration database (UCDB) that contains user-specific data associated with multiple calibration points (CPs) is constructed for a certain number of users. Second, for new users, the candidate calibrated data are retrieved from the UCDB by simply requesting the users to look at a single CP on the screen center (SC). Experimental results show that the proposed REGT method with single-point calibration demonstrates a highly competitive accuracy compared with conventional methods employing multipoint calibration.

Content-preserving video stitching method for multi-camera systems

In this paper, a novel content-preserving video stitching method is proposed. Video stitching has to deal with moving objects in the overlapped area (OA) which often cause problems of structural misalignment or ghost effect. To this end, the proposed method first finds an optimal seam that does not pass through the moving objects, by using the extended dynamic programing (DP) technique based on the energy minimization. Then the content-aware adaptive blending is performed which effectively reduces the color discontinuity while restricting the ghost effect caused by moving objects in the OA. In addition, to reduce the computational complexity, a partial seam-update (PSU) scheme is proposed, in which the seam is re-calculated only for a part of the seam passing through the moving object. Experimental results demonstrate that the proposed method is superior to the conventional ones in terms of both subjective quality and computational complexity. In addition, this method achieves real-time performance on mobile platform, making it applicable to consumer electronic devices.