JPEG XE: enabling interoperable event-based vision systems

Abstract

Event-based imaging, also known as neuromorphic imaging, is a recent visual information modality that relies on bio-inspired sensors that mimic the sensing behavior of the biological retina. Contrary to conventional cameras, where all sensor pixels acquire full images at regular frame intervals, in event cameras each pixel controls the light acquisition asynchronously and independently, triggering an event whenever the light intensity variation exceeds a given threshold. This new type of acquisition yields high temporal resolution, very high dynamic range, low latency, and low power consumption, making event cameras attractive for application scenarios that are challenging for conventional cameras, such as autonomous driving, robotics, AR/VR, and mobile imaging. To unlock this potential, an interoperable and efficient representation of event data is essential. In this context, the Joint Photographic Experts Group (JPEG) has developed the JPEG XE standard, which defines an efficient coding solution for event sequences. This paper presents the motivation, vision, target applications, and key technical aspects of the JPEG XE standard, including its high-level requirements, standardization process, and the core coding system specified in ISO/IEC 26112-1. The design enables simple state-based decoding with flexible message types, achieving high compression efficiency and low latency, making it well suited for heterogeneous event-based vision platforms.