基于像素位移估计的水下湍流退化图像畸形校正算法

Image Deformity Correction Algorithm for Underwater Turbulence Degradation Based on Pixel Displacement Estimation

摘要:

针对由湍流引起的水下畸变退化图像,提出了一种基于估计畸变帧相对于参考帧的像素移位来校正畸变图像的算法。通过自定义度量标准选择参考帧和清晰度较高的输入帧序列,使用像素配准、二维配准和核相关滤波算法对畸变图像进行复原和重建。为验证该算法,分别进行了室内水下湍流模拟系统实验及海洋实地实验,并将实验结果与盲复原算法、凸集投影超分辨率重建算法进行对比。从主观评价上, 图像畸变得到了有效抑制；从客观评价标准来看,指标高于传统的复原和重建算法。实验结果说明,该算法有效提高了水下湍流退化图像的去畸效果,为海洋目标检测的准确作业提供了保证。

Abstract:

Aiming at the degradation image of underwater distortion caused by turbulence, an algorithm based on the pixel shift correction of distorted frames is proposed based on the estimation of the aberration frame relative to the reference frame. Through the custom metric standard, we select the reference frame and a higher definition of the input frame sequence, and use the pixel registration, two-dimensional registration and kernel correlation filtering algorithm to restore and reconstruct the distorted image. In order to verify the algorithm, we carried out indoor underwater turbulence simulation system experiments and ocean field experiments respectively, and compared the experimental results with the blind restoration algorithm and the convex set projection super-resolution reconstruction algorithm. From the subjective evaluation, the image distortion is effectively suppressed. From the objective evaluation standard, the index is higher than the traditional restoration and reconstruction algorithm. Experimental results show that the algorithm effectively improves the dedistortion effect of underwater turbulence degraded image, and provides a guarantee for the accurate operation of marine target detection.