CAJ | 학술논문

针对图像间因存在旋转以及采集图像时光线强度的差异等现象而导致的拼接效果不理想和拼接速度慢的问题，作者提出了一种基于SIFT和改进RANSAC的稳健图像拼接算法，具体包括SIFT特征提取、图像配准以及图像的加权平均融合等步骤。其中，该文重点研究了图像配准这一阶段，此阶段包括图像的初始匹配和精确配准两步，实验证明该拼接算法不仅可以很好的拼接存在平移、旋转、尺度缩放、视角以及光照变化的图像，而且较之传统的RANSAC算法，改进的RANSAC算法迭代次数变少了并且运行时间也减少了，拼接效率得到了明显的提高。
침대도상간인존재선전이급채집도상시광선강도적차이등현상이도치적병접효과불이상화병접속도만적문제，작자제출료일충기우SIFT화개진RANSAC적은건도상병접산법，구체포괄SIFT특정제취、도상배준이급도상적가권평균융합등보취。기중，해문중점연구료도상배준저일계단，차계단포괄도상적초시필배화정학배준량보，실험증명해병접산법불부가이흔호적병접존재평이、선전、척도축방、시각이급광조변화적도상，이차교지전통적RANSAC산법，개진적RANSAC산법질대차수변소료병차운행시간야감소료，병접효솔득도료명현적제고。
Taking into account the problem of poor effect and slow matching in image mosaic caused by phenomena such as the rotation and the differences in light intensity, a steady image stitching algorithm based on SIFT and improved RANSAC is pro?posed, specifically including SIFT feature extraction, image registration and the weighted average image fusion. Among them, the paper focuses on the image registration at this stage, this phase includes an initial matching and accurate alignment. Experimental results show that the proposed image mosaic method can not only yield good results for images with overlap region that existence translation, rotation, image scaling, image viewing angle and illumination changes, and compared with the original RANSAC al?gorithm, the improved method has lower computational complexity, less stitching time, and splicing efficiency has been signifi?cantly improved.