基于FPGA的实时大气湍流图像复原算法及实现

由于大气湍流的影响，远距离光学成像设备获取的图像会出现严重的退化现象，例如几何畸变、运动模糊和离焦模糊等。目前主流的湍流复原算法通常依赖于非刚性配准、重建，或者从长时的视频序列中寻找稀有的“幸运区域”，这些方法均需要庞大的计算量或者提前获得完整视频数据，无法满足实际应用场景的实时性要求。因此，提出一种可在FPGA中实现的实时湍流复原算法。该方法利用大气湍流的随机性，通过连续帧信息对图像进行时间域的滤波，解决了图像几何畸变的问题；然后，将频域的维纳滤波转换为较为容易实现的空间域卷积，解决了图像模糊的问题。实验表明，本文算法不仅满足了实时性要求，同时有效地实现了湍流图像的复原。

FPGA-based real-time atmospheric turbulence image restoration algorithm and its implementation

Due to the influence of atmospheric turbulence, the images obtained by the remote optical imaging equipment will appear serious degradation phenomena, such as geometric distortion, motion blur and defocus blur, etc. At present, the mainstream turbulence restoration algorithms usually rely on the non-rigid registration, reconstruction or finding rare lucky regions from long-term video sequences. These methods require huge calculated amount or complete video data in advance, which cannot meet the real-time requirements of actual application scenarios. Therefore, a real-time turbulence restoration algorithm that could be implemented in field programmable gate array (FPGA) was proposed. This method used the randomness of atmospheric turbulence to filter the image in the time domain through the continuous frames and solved the problem of image geometric distortion. Then, the Wiener filter in the frequency domain was converted into a convolution in the spatial domain, which was easier to implement, and solved the problem of image blur. The experimental results show that the proposed algorithm not only meets the real-time requirements, but also effectively realizes the restoration of turbulence images.