基于虚拟现实技术的图像多特征点匹配拼接方法  

Image multiple feature point matching and stitching methodbased on virtual reality technology

作　　者：董句[1] 陈小丽[1] 刘媛[1] 屠增辉 DONG Ju;CHEN Xiaoli;LIU Yuan;TU Zenghui(Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]华中科技大学,湖北武汉430074

出　　处：《现代电子技术》2023年第19期45-48,共4页Modern Electronics Technique

摘　　要：为解决图像局部区域重叠问题,获取高质量图像,提出基于虚拟现实技术的图像多特征点匹配拼接方法。采用深度图像渲染技术去除原始图像的噪声,减少空洞数量;通过可视化方式确定相邻图像的重叠区域,采用二进制局部特征描述符算法得到重叠区域内的所有特征点;依据夹角反余弦算法粗匹配图像中所有特征点,利用RANSAC算法删掉误匹配点,获得正确匹配点,精准匹配图像多特征点;通过转换矩阵坐标变换正确匹配点实现图像拼接,并通过加权平滑融合方法融合图像拼接过程中产生的痕迹,获取最佳的图像虚拟现实呈现结果。实验结果证明:该方法能够有效实现图像的拼接,消除图像中重叠与错位部分,使拼接后的图像可以达到无缝平滑的效果。An image multiple feature point matching and stitching method based on virtual reality technology is proposed to eliminate overlapping region of images and obtain high⁃quality images.The depth image rendering technology is used to remove the noise of the original image and reduce the number of holes.The overlapping region of adjacent images is determined by means of visualization.All of the feature points in the overlapping region are obtained by the binary local feature descriptor algorithm.All of the feature points in the image are matched roughly according to the included angle arccosine algorithm.The wrong matching points are deleted with RANSAC(random sample consensus)algorithm to obtain the correct matching points,so as to accurately match multiple feature points in the image.The correct matching points are transformed by the transformation matrix coordinates to achieve image stitching.The traces generated in the process of image stitching are fused with the weighted smooth fusion method to obtain the optimal image virtual reality rendering results.The experimental results shows that the method can realize image stitching effectively,eliminate the overlapping and dislocation of the image,and make the stitched image seamless and smooth.

关 键 词：图像拼接 虚拟现实技术 多特征点 粗匹配 精准匹配 深度图像渲染 RANSAC算法 痕迹融合 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP751.1[电子电信—信息与通信工程]