With the rapid development of sensor networks, machine vision faces the problem of storing and computing massive data. The human visual system has a very efficient information sense and computation ability, which has enlightening significance for solving the above problems in machine vision. This review aims to comprehensively summarize the latest advances in bio-inspired image sensors that can be used to improve machine-vision processing efficiency. After briefly introducing the research background, the relevant mechanisms of visual information processing in human visual systems are briefly discussed, including layer-by-layer processing, sparse coding, and neural adaptation. Subsequently, the cases and performance of image sensors corresponding to various bio-inspired mechanisms are introduced. Finally, the challenges and perspectives of implementing bio-inspired image sensors for efficient machine vision are discussed.

CMOS image sensors produced by the existing CMOS manufacturing process usually have difficulty achieving complete charge transfer owing to the introduction of potential barriers or Si/SiO₂ interface state traps in the charge transfer path, which reduces the charge transfer efficiency and image quality. Until now, scholars have only considered mechanisms that limit charge transfer from the perspectives of potential barriers and spill back effect under high illumination condition. However, the existing models have thus far ignored the charge transfer limitation due to Si/SiO₂ interface state traps in the transfer gate channel, particularly under low illumination. Therefore, this paper proposes, for the first time, an analytical model for quantifying the incomplete charge transfer caused by Si/SiO₂ interface state traps in the transfer gate channel under low illumination. This model can predict the variation rules of the number of untransferred charges and charge transfer efficiency when the trap energy level follows Gaussian distribution, exponential distribution and measured distribution. The model was verified with technology computer-aided design simulations, and the results showed that the simulation results exhibit the consistency with the proposed model.

由于紫外光在硅中的穿透深度有限，以及多晶硅栅极对紫外光的吸收，导致传统的硅基CMOS图像传感器在紫外光波段的响应不高。在此，本文选择一种低成本的下转换法来提升CMOS图像传感器的紫外响应能力，采用真空热蒸发法分别在石英衬底和CMOS图像传感器的像敏面上蒸镀了晕苯薄膜，并对薄膜的光学性能、红外光谱、光稳定性和热稳定性进行了研究。实验结果表明，晕苯薄膜能吸收紫外光并发射出500 nm的绿色荧光，可以与CMOS图像传感器的光谱响应峰值很好地匹配；同时，发现晕苯红外吸收光谱的实验值和计算值基本吻合；薄膜在200 ℃温度下退火20 min后，其发射峰的荧光强度保持在原来的95.7%；在280 nm激发波长照射大约60 min后，发光强度呈指数衰减至初始值的64%。采用CMOS单色相机在可见光（400~780 nm）和紫外光（365 nm）下定性分析了薄膜的紫外增强效果，发现蒸镀晕苯薄膜后的CMOS单色相机可以提高对紫外光的灵敏度。

This paper describes a promising route for the exploration and development of 3.0 THz sensing and imaging with FET-based power detectors in a standard 65 nm CMOS process. Based on th plasma-wave theory proposed by Dyakonov and Shur, we designed high-responsivity and low-noise multiple detectors for monitoring a pulse-mode 3.0 THz quantum cascade laser (QCL). Furthermore, we present a fully integrated high-speed 32 × 32-pixel 3.0 THz CMOS image sensor (CIS). The full CIS measures 2.81 × 5.39 mm² and achieves a 423 V/W responsivity (Rv) and a 5.3 nW integral noise equivalent power (NEP) at room temperature. In experiments, we demonstrate a testing speed reaching 319 fps under continuous-wave (CW) illumination of a 3.0 THz QCL. The results indicate that our terahertz CIS has excellent potential in cost-effective and commercial THz imaging and material detection.