An underwater image mosaic method based on multi-scale image fusion and SIFT features

Abstract

The invention discloses an underwater image mosaic method based on multi-scale image fusion and SIFT features. The method comprises the following steps: 1) using an improved white balance algorithm and a CLAHE algorithm based on linear interpolation to perform image enhancement on the underwater image; 2) adopting a double pyramid image fusion method to fuse the enhanced image to obtain an underwater preprocessed image; 3) The underwater preprocessed image is registered through the improved SIFT algorithm; 4) After the image affine transformation matrix is ​​calculated, the final underwater image mosaic is completed using the linear gradient synthesis algorithm. The present invention proves through experiments that this method fully considers the underwater environment and underwater imaging characteristics, and can obviously improve the effect and accuracy of underwater image enhancement, registration and splicing.

Description

technical field [0001] The invention relates to an image mosaic method, in particular to an underwater image mosaic method based on multi-scale image fusion and SIFT features. Background technique [0002] With the maturity of Remote Operated Vehicle (ROV) and Autonomous Underwater Vehicle (AUV) technologies, people can easily obtain underwater optical images through ROV and AUV. However, due to the severe attenuation of light in water, the images and videos obtained by underwater optical imaging equipment have problems such as narrow field of view and low image quality. If you want to obtain images and videos with a wider field of view and more information, you will inevitably reduce their resolution and image quality. In order to solve this problem, image mosaic technology has been produced. In recent years, image stitching technology has developed rapidly, and the most widely used algorithms are SIFT and SURF. Although the SIFT algorithm has strong robustness and good ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above problems and fully consider the characteristics of the deep water environment, the present invention proposes an improved white balance algorithm, which can well solve the serious problem of image color cast when the light is insufficient, so that the image in deep water can also have uniform color distribution and effectively remove The color block can restore the white and gray shades of the image at the same time, and the calculation efficiency is also very high; and the image enhancement algorithm of the multi-scale fusion principle is applied to the underwater preprocessing image, and the improved white balance algorithm and the double pyramid image fusion algorithm are used to improve the image quality. The effect and correct rate of underwater image stitching; and improve the underwater image registration method of SIFT algorithm on the basis of predecessors, and improve the accuracy of underwater image registration and stitching

Images