Correction Methods On Underwater Image Distortion

Along with the 21 st century,people desired to acquire the ocean information more accurately and comprehensivly so they paid more attention to the technology of ocean detection.At present,there are many detection methods for underwater target,such as acoustic wave,radar and photoelectric imaging.Among them,photoelectric imaging detection has become a key research and development technology at home and abroad due to its high efficiency,high resolution and intuitive imaging.However,since the water surface is always perturbed,there are always certain ripples.So the underwater image sequence obtained directly has distortions and it is difficult to obtain specific information.Because the variation of water surface ripples is so intense,there is still some difficulty in using adaptive optics wavefront correction method for submarine target imaging.To this end,many scholars carried out recovery methods based on image processing and lucky region is the most representative.The main research of the thesis is the underwater distortion image correction algorithm based on lucky region,and the simulation experiments are carried out on software platform such as Matlab.The research could be divided into two parts.First,we propose a lucky region algorithm based on structure tensor and structural similarity index.The new algorithm not only corrects the distortion but also improve the contrast.The results show the recovery image's view is better than the traditional algorithm and the PSNR is improved from 14.4 to 15.1.Second,taking into account the limitations of the lucky region algorithm,this paper presents a method which combines motion estimation and lucky region.Under normal circumstances,the temporal mean image of underwater image sequence has a relatively low distortion.Each frame is registered to mean image to get motion vector used a free transform model based on B-spline.Then,with the help of motion vector,each frame is compensated using interpolation.Then a new mean image is calculated with the compensated frames and the process is repeated.After iteration,the distortion of image sequence is reduced significantly and then lucky region algorithm is used to calculate a high-quality reconstructed still image.The experimental results show that the proposed algorithm has greatly reduced the distortion compared with using lucky region only and the PSNR is improved from 15.1 to 17.9.Compared with traditional methods,the algorithm has advantages both in subjective and SSIM evaluation.Our results can provide some reference value for the resarch of underwater distortion image restoration.