摘要
视频对象的分割在基于内容的视频编码及视频检索中占有重要的地位。在快速准确的原则上,利用视频序列中的时域及空域信息,提出了一种分割运动对象的仿真方法。首先利用二次帧差求交集的方法求出图像帧的运动区域,然后利用形态梯度算子得到图像帧的边缘信息,最后时空结合并填充就可以分割出运动对象。实验证明分割效果较好并节省了处理时间。
Video object segmentation is important in content-based video encoding and video retrieval. Based on the rules of speediness and exactness, the method of moving object segmentation is proposed with temporal and spacial information. We obtain moving regions by the intersection of three twice-frame-difference images, then get boundaries of the frame by morphological gradient operation. At last, integrating temporal information with spacial information and filling with original values gets extracted moving object. The experiments show that the method can get a better result with less runtime.
出处
《中国体视学与图像分析》
2005年第2期104-107,共4页
Chinese Journal of Stereology and Image Analysis
基金
山东省自然科学基金(Y2001G05)
关键词
运动区域
运动对象
二次帧差
形态梯度算子
moving regions
moving objects
twice-frame-difference
morphological gradient operation
