电子设计工程
電子設計工程
전자설계공정
ELECTRONIC DESIGN ENGINEERING
2015年
12期
116-119
,共4页
关键帧%SIFT特征点%单映矩阵%图像拼接%RANSAC鲁棒估计算法
關鍵幀%SIFT特徵點%單映矩陣%圖像拼接%RANSAC魯棒估計算法
관건정%SIFT특정점%단영구진%도상병접%RANSAC로봉고계산법
key frame%SIFT feature%homography matrix%image mosaic%RANSAC robust estimation algorithm
为消除视频序列之间的冗余信息，以简单的摘要形式表达视频的主要内容，提出了一种基于视频序列的图像拼接方法。首先，采用改进的帧间聚类算法提取视频的关键帧；其次，利用SIFT算法提取关键帧的特征点，采用最近邻算法进行特征点匹配，通过引导互匹配法和投票过滤法提高匹配精度；再次通过RANSAC鲁棒估计算法得到所选帧间的单映矩阵，并使用LM非线性迭代算法对单映矩阵进行精炼。最后，利用级联单映矩阵结合加权融合算法实现了视频序列的无缝拼接，实验效果较为理想。
為消除視頻序列之間的冗餘信息，以簡單的摘要形式錶達視頻的主要內容，提齣瞭一種基于視頻序列的圖像拼接方法。首先，採用改進的幀間聚類算法提取視頻的關鍵幀；其次，利用SIFT算法提取關鍵幀的特徵點，採用最近鄰算法進行特徵點匹配，通過引導互匹配法和投票過濾法提高匹配精度；再次通過RANSAC魯棒估計算法得到所選幀間的單映矩陣，併使用LM非線性迭代算法對單映矩陣進行精煉。最後，利用級聯單映矩陣結閤加權融閤算法實現瞭視頻序列的無縫拼接，實驗效果較為理想。
위소제시빈서렬지간적용여신식，이간단적적요형식표체시빈적주요내용，제출료일충기우시빈서렬적도상병접방법。수선，채용개진적정간취류산법제취시빈적관건정；기차，이용SIFT산법제취관건정적특정점，채용최근린산법진행특정점필배，통과인도호필배법화투표과려법제고필배정도；재차통과RANSAC로봉고계산법득도소선정간적단영구진，병사용LM비선성질대산법대단영구진진행정련。최후，이용급련단영구진결합가권융합산법실현료시빈서렬적무봉병접，실험효과교위이상。
In order to eliminate the redundant information between video sequences, express the main content of the video in a simple abstract form, an image mosaic method based on video sequence was proposed. First of all, the key frames were extracted using the improved inter frame clustering algorithm. Secondly, the feature of key frames were extracted by using the SIFT algorithm, the nearest neighbor algorithm is used to match the feature points, guided complementary matching and voting filter is used to improve the matching accuracy. Then with the RANSAC robust estimation algorithm to get the selected frames' homography matrix. The LM nonlinear iteration algorithm was used to refine the homography matrix. Finally, cascade homography matrix and weighted fusion algorithm is used to realize the seamless splicing of video sequences. The experiment demonstrates this method is efficient.
