【摘要】 提出了一种结合视差和边缘信息进行立体视频对象的分割方法。针对传统分水岭算法过分割的问题,提出了改进的算法。采用基于自适应权值的立体匹配方法获得可靠的视差图,并利用改进分水岭算法进行视差图分割,获得初始的视频对象区域,随后对初始视频对象进行边缘检测,通过提取其轮廓获得准确的对象,最后提出基于区域的对象跟踪方法,完成后续帧视频对象的分割。实验结果表明,本文提出的方法可以获得良好的分割结果。更多还原

【Abstract】 A method for stereoscopic video object segmentation based on disparity and edge information was presented.In order to solve the over-segmentation problem of traditional watershed algorithm,an improved version was proposed.Disparity map was acquired using adaptive weight stereo matching approach.Then it was segmented by improved watershed algorithm to get initial video objects.For each coarse object,edge detection was implemented to obtain accurate contours of the object.Using the extracted objects,an efficient motion tracking algorithm based on region was performed to track the video objects of the rest frames.The experimental results show that the proposed method can segment video objects successfully.更多还原