CAJ | 학술논문

提出了一种结合空域和频域进行全景图像序列自动拼接融合的优化算法。首先利用改进的相位相关法对全景序列图像进行自动排序并确定重叠区域，根据重叠区域像素均值进行图像整体亮度差异自动调整以降低角点的误匹配率；然后使用改进的Harris算子在空间域提取图像角点(无需人工设定阈值)，通过双向最大互相关系数匹配获得初始特征点对，并用RANSAC算法实现精确匹配；最后采用非线性平滑算法对图像重叠区域进行融合处理。实验结果表明，该优化算法排序过程简单有效，特征点提取匹配过程的成功率和效率都较现有算法有很大提高，拼接的图像清晰度高，具有较高的稳健性和拼接精度。
제출료일충결합공역화빈역진행전경도상서렬자동병접융합적우화산법。수선이용개진적상위상관법대전경서렬도상진행자동배서병학정중첩구역，근거중첩구역상소균치진행도상정체량도차이자동조정이강저각점적오필배솔；연후사용개진적Harris산자재공간역제취도상각점(무수인공설정역치)，통과쌍향최대호상관계수필배획득초시특정점대，병용RANSAC산법실현정학필배；최후채용비선성평활산법대도상중첩구역진행융합처리。실험결과표명，해우화산법배서과정간단유효，특정점제취필배과정적성공솔화효솔도교현유산법유흔대제고，병접적도상청석도고，구유교고적은건성화병접정도。
An optimized algorithm of automatic panoramic image mosaic was presented based on time and frequency domains. Firstly, the improved phase correlation method was used to sort the panoramic image sequence automatically and then determine the overlap region. In order to reduce the false match rate of the corner, the brightness of the overall image was adjusted automatically according to the overlap region's pixel mean. Then, an improved Harris operator was adopted to extract corners (with no need to set threshold value manually), the initial feature points pairs were obtained by the bidirectional greatest correlative coefficient and the false feature point pairs were rejected by RANSAC algorithm. Finally, the nonlinear smoothing algorithm was adopted to fuse the overlap regions of the image. The experimental results show that optimization algorithm is simple and effective in the sorting process. Compared with the existing algorithm, the optimization algorithm proposed at present makes a great progress in the success rate and the efficiency in the matching and extracting process of feature points. The stitched image is with great clarity, soundness and stitching accuracy.