Research On Multi-scale Image Fusion And Stitching Algorithm For Underwater Environment

Image stitching,an image processing technology,has been applied to seabed exploration in recent years.It is the process of resynthesizing and fusing multiple images with overlapping parts,and integrating them into a wide-angle,wide-range panoramic image.In this paper,image stitching is mainly applied to underwater images which have severe color bias,high noise and low contrast,so underwater image preprocessing has a great impact on the accuracy of subsequent image registration.According to the characteristics of underwater environment,a multi-scale image fusion algorithm is proposed to preprocess underwater images,and an improved SIFT algorithm is proposed to register and stitch underwater images.The underwater image stitching algorithm proposed in this paper includes three parts: underwater image preprocessing,underwater image registration and underwater image synthesis.For underwater image preprocessing,a multi-scale image fusion algorithm is proposed in this paper according to the characteristics of underwater environment: Firstly the improved white balance algorithm is proposed to correct color cast,then use the bilateral filtering to remove noise and CLAHE based on linear interpolation to enhance image contrast,thus get one image for image fusion.Secondly,the standard weight image of the preprocessed image is calculated and another image to be fused is obtained.Finally,this paper uses the double pyramid algorithm to merge the two images to get the final underwater preprocessing image.Experiments verify that the multiscale image fusion algorithm proposed in this paper builds models specifically for underwater imaging characteristics,and can well repair images with severe color bias in deep waters.The algorithm is fast and effective.For underwater image registration,an improved SIFT algorithm is proposed in this paper.Because the brightness and contrast of underwater images are generally low,an adaptive threshold method is proposed to screen key points and expand the extraction range of key points.In view of the large amount of underwater image noise,Canny algorithm is used to calculate the gradient and size of key points in this paper,so that more details of the image could be preserved while smoothing the noise.When matching key points,the Hausdorff distance and BBF algorithm are used for rough matching of key points,and then RANSAC algorithm is used for further purification.Experiments verify that the improved SIFT algorithm proposed in this paper can quickly obtain the successfully matching key points and significantly improve the matching accuracy.Finally,image synthesis: Firstly,the transformation matrix is calculated.After the bilinear interpolation of the registered images,the images are synthesized to obtain the final stitching image.Through experiments,this paper verifies that the proposed multi-scale image fusion algorithm and improved SIFT algorithm are suitable for the characteristics of underwater environment,which can significantly improve the quality and visual effect of underwater images,and improve the effect and accuracy of underwater image registration and stitching.