It is challenging to removing atmospheric turbulence-induced geometric distortion and blurry degradation simultaneously. The improved principal component analysis (i-PCA) method is proposed in this paper. The image set selection strategy is implemented by the Euclidean distance between each sub-module image and its corresponding reference image to remove large geometric deformations and blurs directly. The eigenvector representing the original image data is uniquely determined by combining the maximum correlation with the original image and the maximum variance of the corresponding principal component for resolving the direction issue of the eigenvector. The algorithm is tested using synthetic and real turbulence-distorted images. The experimental results show that the i-PCA method proposed can effectively alleviate the atmospheric turbulence effects and significantly enhance visual quality, promising for its application to turbulence-distorted image.

同时消除大气湍流引起的几何失真和模糊退化具有挑战性。本文提出了改进的主成分分析（i-PCA）方法。图像集选择策略是通过每个子模块图像与其对应的参考图像之间的欧几里德距离来实现的，以直接去除大的几何变形和模糊。通过与原始图像的最大相关性和对应主成分的最大方差相结合，唯一确定代表原始图像数据的特征向量，以解决特征向量的方向问题。该算法使用合成和真实的湍流失真图像进行测试。