摘要
充分考虑水下环境和水下成像的特点,将多尺度融合的图像增强算法应用于水下配准算法的预处理图像中,修复深水域偏色严重的图像。用改进SIFT算法进行特征提取,采用自适应阈值法筛选关键点,扩大关键点提取范围;用Canny算法计算关键点的梯度和大小,平滑噪声的同时也可以保留图像更多细节;使用平均Hausdorff距离和BBF最邻近查询法对关键点进行粗匹配,再用RANSAC进行进一步提纯;计算得到变换矩阵后输出最终拼接图像。实验验证了该算法适合水下环境特点,可以提升水下图像配准和拼接的效果和准确率。
Considering the characteristics of underwater environment and underwater imaging,the image enhancement algorithm based on multi-scale fusion was applied to the pre-processing images of underwater registration algorithm to repair the images with severe color deviation in deep water.The improved SIFT algorithm was used to extract features,and the adaptive threshold was adopted to screen key points and expand the range of key points extraction.The Canny algorithm was used to calculate the gradient and size of the key points,and more details of the image could be retained while smoothing the noise.The average Hausdorff distance and the nearest query method of BBF were used to roughly match the key points,and then RANSAC was used for further purification.After calculating the transformation matrix,the final stitching image was output.The experimental results show that this algorithm is suitable for the characteristics of underwater environment,and it can improve the effect and accuracy of underwater image registration and stitching.
作者
王昕平
张森林
刘妹琴
樊臻
Wang Xinping;Zhang Senlin;Liu Meiqin;Fan Zhen(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,Zhejiang,China)
出处
《计算机应用与软件》
北大核心
2021年第5期213-217,230,共6页
Computer Applications and Software
基金
国家自然科学基金-浙江两化融合联合基金项目(U1809212)
浙江省科技厅重点研发计划项目(2018C03030)。
