The motion of the object or the infrared (IR) imaging system during the integration time causes blurring of the IR image. This study covers real-time field programmable gate array (FPGA)-based deblurring for IR detectors, and an inertial measurement unit (IMU) was used to quantify the blur caused by the IR detector movement. Point spread function for each pixel was calculated using the angular position data of the IR detector obtained from IMU. Both spatially invariant and spatially variant blur cases can be modeled for the IR detector motion. After the quantification, the spatially invariant-type blur was eliminated using a Wiener filter-based deblurring algorithm. Deblurring algorithm was implemented in the Xilinx system generator environment directly using FPGA IP cores. The simulation results in the Xilinx system generator environment indicate that the proposed image deblurring method is real-time applicable, and it reduces the processing time of a single frame to 4 ms. For the implementation of 2D-fast Fourier transform design in FPGA using the corner turn matrix method, memory management is the most critical factor influencing the speed. The real-time deblurring solution given herein has the potential to be used in IR cameras on the moving platforms to increase the performance and robustness in systems such as object tracking and visual navigation.

中文翻译：

---

使用红外探测器角位置的基于FPGA的红外图像去模糊

在积分时间内物体或红外 (IR) 成像系统的运动会导致红外图像模糊。这项研究涵盖了基于实时现场可编程门阵列 (FPGA) 的红外探测器去模糊，并使用惯性测量单元 (IMU) 来量化由红外探测器运动引起的模糊。使用从 IMU 获得的红外探测器的角位置数据计算每个像素的点扩散函数。空间不变和空间变化的模糊情况都可以为 IR 检测器运动建模。量化后，使用基于维纳滤波器的去模糊算法消除空间不变型模糊。去模糊算法是直接使用 FPGA IP 核在 Xilinx 系统生成器环境中实现的。在 Xilinx 系统生成器环境中的仿真结果表明，所提出的图像去模糊方法是实时适用的，并将单帧的处理时间减少到 4 ms。对于使用转角矩阵法在FPGA中实现二维快速傅里叶变换设计，内存管理是影响速度的最关键因素。本文给出的实时去模糊解决方案有可能用于移动平台上的红外摄像机，以提高系统的性能和鲁棒性，例如对象跟踪和视觉导航。内存管理是影响速度的最关键因素。本文给出的实时去模糊解决方案有可能用于移动平台上的红外摄像机，以提高系统的性能和鲁棒性，例如对象跟踪和视觉导航。内存管理是影响速度的最关键因素。本文给出的实时去模糊解决方案有可能用于移动平台上的红外摄像机，以提高系统的性能和鲁棒性，例如对象跟踪和视觉导航。