Research On Fast Mosaic Algorithm Of Underwater Image Based On SURF

Rational development of marine resources,effectively protect marine biodiversity is of great significance to humans.Because the field of vision of a single underwater image is limited,it can not fully show the underwater appearance,so a fast underwater image stitching algorithm based on SURF(Speeded-Up Robust Feature)is proposed in this paper.The fast stitching technology of underwater image is the process of quickly stitching the input underwater image into a complete image,which is mainly composed of underwater image enhancement,image matching and image fusion.Firstly,this paper introduces the basic principle of quickly extracting SURF feature points from underwater images.The fast detection of feature points is realized by using integral image and discrete box filter,and the stable construction of feature descriptor is completed with the help of Haar wavelet filter.Secondly,in this paper,a fast underwater image enhancement method based on multi-scale fusion is proposed in this paper.An underwater image white balance method based on color compensation is established to quickly solve the problem of image color deviation caused by color imbalance.Aiming at the problems of low contrast and unclear details in the white balance image,the white balance image is processed by Gamma correction and sharpening operation,and the pixels of the sharpened image are adjusted combined with the stretching of the histogram obeying Rayleigh distribution.Then the method of multi-scale fusion is used to improve the underwater image quality as a whole.The experimental results show that the underwater enhanced image obtained by this method has a good effect.Then,the fast stitching of the two images is completed on the underwater enhanced image.In this paper,the nearest neighbor ratio method and RANSAC(Random Sample Consensus)method are used to initially match and accurately match the extracted SURF feature points,and the exact matching point pair is used to obtain the coordinate transformation matrix to achieve accurate underwater image registration.The results show that this method can obtain underwater registration images quickly and accurately.In addition,by combining the characteristics of the underwater image enhancement method,the gradual in-and-out fusion method can be used to fuse the overlapping regions of underwater images quickly and effectively.Finally,the automatic panoramic stitching of disordered underwater images is realized.In order to keep the visual consistency of the panoramic image,the input underwater images are projected to the cylindrical coordinate system,and the probability model is used to solve the difficult problem of disordered underwater image matching.Then the strategy of grouped image mosaic is used to reduce the cumulative error in the process of panoramic mosaic.The results show that this method can stitch underwater images automatically,accurately and quickly.