Robust Iris Centre Localisation for Assistive Eye-Gaze Tracking

Nipun Sandamal Ranasekara Pathiranage,Stefania Cristina,Kenneth P. Camilleri

2024-11-08

Abstract:In this research work, we address the problem of robust iris centre localisation in unconstrained conditions as a core component of our eye-gaze tracking platform. We investigate the application of U-Net variants for segmentation-based and regression-based approaches to improve our iris centre localisation, which was previously based on Bayes' classification. The achieved results are comparable to or better than the state-of-the-art, offering a drastic improvement over those achieved by the Bayes' classifier, and without sacrificing the real-time performance of our eye-gaze tracking platform.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is **robust iris center localization under unconstrained conditions**, which is a key component of its eye - tracking platform. Specifically, the research aims to improve the accuracy and real - time performance of iris center localization to overcome the limitations of existing methods (such as Bayesian classifiers) in complex real - world environments.

### Problem Background
1. **Limitations of Existing Methods**:
- Early work used Bayesian classifiers for iris region segmentation and estimated the iris center by calculating the centroid of the binarized blob. This method performs well when the illumination is sufficient and the iris is visible, but in practical applications, due to factors such as eyelid occlusion and shadows, the Bayesian classifier is prone to mis - segment the iris region, resulting in inaccurate iris center estimation.

2. **Application of Deep Learning**:
- In recent years, convolutional neural networks (CNNs) have been widely applied to iris segmentation tasks, but these models are often complex and computationally expensive.
- Some studies formulate the iris / pupil center localization problem as a heatmap regression problem, that is, generate a heatmap and find the "hottest" point as the iris center coordinates, but these methods perform poorly when dealing with eye occlusions, reflections and shadows.

### Research Objectives
To improve the accuracy of iris center localization while maintaining the real - time performance of the system, the authors propose two methods based on U - Net variants:
1. **Segmentation - Based Method**: Use the U2 - Net model for iris region segmentation, and then locate the iris center by calculating the centroid of the segmentation result.
2. **Heatmap Regression - Based Method**: Use the modified U - Net model to generate a heatmap, and locate the iris center by finding the "hottest" point in the heatmap.

### Main Contributions
- **Improved Accuracy**: Experimental results show that these two methods perform better than or at least on a par with the existing state - of - the - art methods on multiple datasets.
- **Real - Time Performance**: The proposed models can ensure the real - time performance of the system while maintaining high precision, which is suitable for assisting eye - tracking platforms.

### Conclusion
By introducing U - Net variant models, this research significantly improves the robustness and accuracy of iris center localization, especially in complex real - world environments. Future work will further integrate these models into real - time eye - tracking platforms and test them in assistive technology scenarios.