摘要
为提升AR动态图像轮廓特征匹配准确性,扩展图像拼接技术的应用领域,满足信息背景下的计算机视觉与图像处理需求,提出一种AR动态图像轮廓特征匹配拼接方法。通过滤波处理、平滑与边缘锐化以及几何校正阶段,完成图像预处理,划分图像为左边缘及右边缘,搜索不间断边缘点上的全部轮廓点,基于方向性约束与灰度相关性约束获取最优轮廓匹配点,利用特征向量信息实现匹配检验与轮廓特征匹配,根据图像梯度场,引导插值拼接区域,转换图像拼接问题为最小化目标函数问题,采用泊松融合完成AR动态图像轮廓特征匹配拼接。从无标准参考与有标准参考两种图像角度,使用不同的评估指标来评价拼接图像质量,仿真结果表明,所提方法具有明显的配准优越性与较好的拼接效果,有效性与适用性相对理想,能够满足信息背景下图像处理需求。
It is necessary to improve the accuracy of AR dynamic image contour feature matching and expand the application field of image mosaic technology for meeting the needs of computer vision and image processing under the background of information. Therefore, this paper proposes an AR dynamic image contour feature matching mosaic method. Based on filtering, smoothing, edge sharpening and geometric correction, an image was preprocessed to divide into left edge and right edge. All contour points on continuous edge points were searched. The optimal contour matching points were obtained by directional constraint and gray correlation constraint. The gradient field was used to guide the interpolation stitching area, and the image stitching was converted to the minimum objective function. Finally, the contour feature matching stitching of AR dynamic image was achieved based on Poisson fusion. The simulation results show that this method has obvious registration superiority and excellent mosaic effect, ideal validity and applicability, and meets the needs of image processing in the future information background.
作者
谭建梅
黄隽
TAN Jian-mei;HUANG Jun(Shanxi Datong University,Datong Shanxi 037009,China;South Central University for Nationalities,Wuhan Hubei 430000,China)
出处
《计算机仿真》
北大核心
2021年第6期138-141,272,共5页
Computer Simulation
基金
2020年山西省哲学社会科学规划课题(2020YY175)。
关键词
信息背景
动态图像
轮廓特征匹配
图像拼接
图像灰度
Information background
Augmented Reality
Dynamic image
Contour feature matching
Image mosaic
Image grayscale
