Event Camera Calibration Using Vanishing Points

Camera calibration is a fundamental process for 3D computer vision. While well-established for traditional framebased cameras, calibrating novel event-based cameras remains a key challenge, as conventional methods are not directly applicable due to the distinct nature of event data. This paper introduces the first framework for calibrating event-based cameras using vanishing points. Our method robustly estimates the camera’s intrinsic parameters from a single view of a common prismatic object (e.g., a box or a book), eliminating the need for any dedicated calibration patterns.We validate our approach through two distinct experiments: one using raw event data from a custom-designed cube with blinking LEDs, and another using images reconstructed from event streams recorded in scenes containing off-the-shelf objects commonly found in man-made environments. Comparative analysis demonstrates that our method achieves high accuracy, with results closely matching those from widely used calibration tools. The proposed framework offers a fast, practical, and tool-independent solution for event camera calibration. The dataset and source code are made publicly available at https://github.com/SezginDulkadir/ECCVP.