The main contribution of this paper is to provide an easy‐to‐implement and low‐cost gaze tracking system for near‐eye display, which can meet the needs of application scenarios such as near‐eye display interaction and foveated rendering. The hardware of the system is an infrared camera combining with an annular infrared light source of 850nm to achieve image acquisition. The software algorithm is based on the pupil segmentation algorithm of sliding window with adaptive threshold, which can achieve the precise pupil segmentation, the pupil ellipse fitting method that can achieve the precise pupil positioning, and the polynomial model that can establish the mapping between the pupil center and the fixation point. Consider the real application environment, the eye position compensation method based on the canthus point is adopted to achieve a more stable gaze calculation which completed a more perfect VR gaze‐tracking system. For the experimental results, the average error for the system is 0.55° in the horizontal direction and 0.63° in the vertical direction, the latency on the Intel I7‐6700HQ CPU is about 3.5 ms, which indicates that the system can achieve the gaze‐tracking calculation with high speed and precision that can be used in VR.
For safety reasons, the current car's A‐pillars are generally designed to be relatively wide, causing a large visual blind area and resulting in a great safety hazard in driving process[1]. This paper presents a visual display system for the A‐pillar to eliminate the blind area, which combines face detection, pupil detection, gaze calculation and image extraction module, and achieves displaying effect of scene moving with people and seamless splicing. Only the head position is calculated in traditional methods because the head position cannot represent the gaze direction, causing a large spliced error. In this paper, when calculating the spliced image, gaze calculation algorithm is added to the visual system, which greatly improves the calculation accuracy of the spliced image, reaching 1.6°, and improves final splicing performance.
Compared with traditional methods, edge-based interpolation algorithm can generate high-resolution images with rare edge artifacts. However complex textured regions produce morphological distortion at the same time, resulting in a HD image looks unnatural. In this paper, we analyze the disadvantages of the best performing edge-based interpolation (DCCI) algorithm and propose a novel adaptive edge-based interpolation algorithm, which use both 2-D texture entropy and directional gradient to eliminate the distortion. It turns out that this method can effectively improve the image quality.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.