CAJ | 학술논문

针对稳像过程中的局部运动干扰, 提出了一种基于特征块角点检测的电子稳像算法。该算法采用SUSAN算子, 对均匀分布的大尺寸特征块进行检测, 并构造特征块及对应检索区域角点集合。通过计算基于角点响应函数的修正Hausdorff距离实现背景局部遮挡下的特征块配准, 进而通过运动矢量一致性检测得到有效运动矢量。实验结果表明, 该方法对实测抖动视频序列的平均峰值信噪比在稳像后提高了19. 83 dB, 对局部运动干扰下的抖动图像序列具有较好的稳像性能。
침대은상과정중적국부운동간우, 제출료일충기우특정괴각점검측적전자은상산법。해산법채용SUSAN산자, 대균균분포적대척촌특정괴진행검측, 병구조특정괴급대응검색구역각점집합。통과계산기우각점향응함수적수정Hausdorff거리실현배경국부차당하적특정괴배준, 진이통과운동시량일치성검측득도유효운동시량。실험결과표명, 해방법대실측두동시빈서렬적평균봉치신조비재은상후제고료19. 83 dB, 대국부운동간우하적두동도상서렬구유교호적은상성능。
This paper proposed a video stabilization algorithm based on feature block corner detection for suppressing local motion interference. It detected corners of uniform distribution feature blocks by using SUSAN operator. It realized feature block registration under partial occlusion by calculating the similarity of the two corner sets with partial mean Hausdorff distance modified with corner response function. It chose the effective motion vectors by making consistency checking. The expe-rimental results show that the average PSNR is raised 19. 83 dB after stabilization. The proposed algorithm can stabilize image sequence with local motion interference.