移动AR+VR支持下全景图像拼接均匀性校正方法  

Uniformity Correction Method of Panoramic Image Mosaic Supported by Mobile AR+VR

作　　者：赵松[1] 王勐 ZHAO Song;WANG Meng(Nanjing Normal University Zhongbei College,Danyang Jiangsu 212300,China)

机构地区：[1]南京师范大学中北学院,江苏丹阳212300

出　　处：《计算机仿真》2021年第11期189-192,398,共5页Computer Simulation

摘　　要：采用传统方法对全景拼接图像校正时,无法检测到全景图像中的重叠区域,导致全景图像结构相似度评分SSIM偏低、边缘差分谱评分DoEM低且校正误差较大。为解决上述问题,提出移动AR+VR支持下全景图像拼接均匀性校正方法。采用相位相关算法在移动AR+VR支持下检测全景图像重叠区域。确定全景图像的畸变中心坐标和畸变系数,将图像畸变区域投影到地图坐标中,通过亮度像素值校正镜头畸变和斜视变形,完成图像的均匀性校正。实验结果表明,所提方法的结构相似度评分SSIM均大于0.93,边缘差分谱评分DoEM最高可达0.9914,且扭曲度均低于10%,说明所提方法累计误差小,实验数据验证了所提方法具有明显应用优势。In the process of panoramic mosaic image correction, the image overlapping area detected by traditional methods is blurred, resulting in low panoramic image structure similarity score(SSIM), low edge difference spectrum score(DOEM) and large correction error. Therefore, this paper reports a panoramic image mosaic uniformity correction method based on mobile AR+VR support. Firstly, the phase correlation algorithm was introduced to detect the overlapping region of panoramic image. Secondly, the distortion center coordinates and distortion coefficients of the panoramic image were obtained. Then, the image distortion region was projected into the map coordinates. Finally, lens distortion and squint distortion were corrected according to the brightness pixel value, thus completing the uniformity correction of the image. The experimental results show that the method has low cumulative error(SSIM > 0.93, DOEM=0.9914) and low distortion(< 10%).

关 键 词：全景拼接图像 相位相关算法 直线透视投影不变性原理 均匀性校正 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]