时滞与积分
微测辐射热计
动态范围
像素
高动态范围
计算机科学
图像传感器
高动态范围成像
控制重构
宽动态范围
帧速率
电子工程
探测器
计算机视觉
人工智能
集成电路
CMOS芯片
图像质量
动态范围压缩
图像处理
实时计算
卷帘
红外线的
计算机硬件
数字图像相关
工程类
目标检测
图像分辨率
动态需求
炸薯条
自动增益控制
作者
Ke Li,Yingjian Hao,Yina Ma,Ke Li,Jiqing Zhang,Yao Li
摘要
The growing demand for high dynamic range (HDR) imaging in modern intelligent security systems and industrial inspection has revealed a fundamental limitation of conventional infrared focal plane arrays (IRFPAs): their global-shutter-based integration time configurations fails to simultaneously resolve high-intensity and low-intensity regions within a single frame, leading to persistent highlight saturation and low-light invisibility. Although conventional HDR algorithms merge multi-frame images with varied integration times to extend dynamic range, such methods suffer from inherent latency. The inter-frame temporal disparities induce motion blur/ghosting artifacts and constrain the peak frame rate of the detector. To address the critical trade-off between dynamic range and real-time imaging performance in single-exposure infrared imaging, this work proposes a novel image sensor design featuring per-pixel adjustable integration time, implemented through a direct-injection pixel architecture with embedded functional modules. Each pixel integrates in-pixel memory cells, timing counters, and digital comparator, enabling individual exposure configuration during a single integration cycle via SPI interface. Fabricated in a standard 0.13 μm CMOS process, the readout integrated circuit (ROIC) incorporates a 640×512 pixel array with 15-μm pixel pitch. Experimental characterization demonstrates: (1) linear correlation between output response and integration time across a wide range of irradiation intensities; (2) a 47.2 dB dynamic range enhancement versus conventional infrared detectors when employing multi-tiered integration time configurations with image fusion algorithms. The proposed circuit preserves both highlight and shadow details under extremely high illumination conditions. Moreover, its per-pixel programmability supports on-chip real-time reconfiguration for adaptive optimization of imaging quality based on incident radiation intensity, significantly advancing applications requiring simultaneous high dynamic range and instantaneous response.
科研通智能强力驱动
Strongly Powered by AbleSci AI