NEOSTI - a neuromorphic electronic-opto spatial-temporal hybrid image sensor - Open Access - 01.12.2026 - Article Abstract Introduction The human visual system is highly efficient in collecting and processing of the visible information. As a bionic device, the image sensor establishes a bridge between the real world in optical domain and artificial intelligence in electronic domain, but current image sensor technology falls short in meeting the requirement of real-time, low-power processing capability at the edge, which is critical for embodied intelligence such as intelligent robots, fully autonomous driving, and drones that require effective interaction with dynamic environments. Different processing technologies in either electronic domain or optical domain have been proposed in literature to break the bottleneck, but there is always a trade-off between processing performance and power/volume consumption. Traditional artificial vision system relies on image sensors coupled with cloud processing or Artificial Intelligence (AI) processors, featuring higher image resolution, higher frame rate, and higher accuracy across various image processing tasks than human visual system, but paying for a several orders higher power consumption than the human visual system. In addition, the sub-second wireless communication latency1 between the sensor end and the computational cloud cannot meet the milliseconds to sub-milliseconds response time requirement risen from the embodied intelligence scenario. Anzeige In order to improve efficiency of the traditional artificial vision system, Processing-in-Sensor (PIS)2‐9 and Processing-Near-Sensor (PNS)10‐12 technologies have been introduced into the raw intensity acquisition process. PIS/PNS technology mitigates the delay and power consumption associated with redundant data transmission workload by integrating the sensing and computing components, enhances data processing efficiency, as well as reduces the volume of data transmitted to the host through parallel processing at column or pixel levels. This approach aligns more closely with the operational principles of the visual systems observed in humans and various creatures. Nevertheless,