Pupil detection plays a key role in eye and gaze video-based tracking algorithms. Various algorithms have been proposed through the years in order to improve the performances or the robustness in real-world scenarios. However, the development of an algorithm which excels in both execution time and pupil detection precision is still an open challenge. This paper presents a novel, feature-based eye-tracking algorithm for pupil detection. Morphological operators are used to remove corneal reflections and to reduce noise in the pupil area prior to the pupil detection step: this solution allows to significantly reduce the computational overhead without lowering the tracking precision. Moreover, a shape validation step is performed after the elliptical fitting and, if the elliptical shape is not detected properly, a set of additional steps is performed to improve the pupil estimation. The proposed solution, Pupil Detection after Isolation and Fitting (PDIF), has been compared with other state-of-the-art tracking algorithms that use morphological operations such as ElSe (Ellipse Selection) and ExCuSe (Exclusive Curve Selector) to evaluate both speed and robustness; the proposed algorithm has been tested over numerous datasets offering different pupil detection challenges. Obtained results show how PDIF provides comparable tracking precision at a significantly lower computational cost compared to ElSe and ExCuSe.
PDIF: Pupil Detection After Isolation and Fitting / Manuri, Federico; Sanna, Andrea; Petrucci, Christian Pio. - In: IEEE ACCESS. - ISSN 2169-3536. - ELETTRONICO. - 8(2020), pp. 30826-30837.
Titolo: | PDIF: Pupil Detection After Isolation and Fitting |
Autori: | |
Data di pubblicazione: | 2020 |
Rivista: | |
Digital Object Identifier (DOI): | http://dx.doi.org/10.1109/ACCESS.2020.2973005 |
Appare nelle tipologie: | 1.1 Articolo in rivista |
File in questo prodotto:
File | Descrizione | Tipologia | Licenza | |
---|---|---|---|---|
08990093.pdf | Articolo | 2a Post-print versione editoriale / Version of Record | ![]() | Visibile a tuttiVisualizza/Apri |
http://hdl.handle.net/11583/2797356