图像，视频的分割技术研究

发布时间：2018-03-01 00:25

本文关键词： 图像分割视频分割多分辨率分割方法基于视频体的分割　出处：《上海交通大学》2010年硕士论文　论文类型：学位论文

【摘要】：近几十年来,数字图像,视频的处理技术一直受到广泛关注。不仅是专业的影视、动漫企业,更多的非专业用户希望对自己拍摄的媒体素材进行编辑处理工作。图像,视频的分割技术在素材编辑中起着重要的作用。用户从素材中提取的前景对象可以被运用在遮片,融合等后期处理技术中,从而达到素材高效复用的目的。因此研究图像,视频的分割技术并且开发出有良好用户交互性的素材编辑工具有着重要意义。本文对图像分割和视频分割技术进行了研究。现有的图像分割算法已非常成熟,但随着数字高清图像,视频捕获设备的发展,产生了大量的高清数字媒体素材。传统的分割技术难以高效运用在这些大数据量的素材上。多分辨率的图像分割方法能减少处理节点的数量,提高算法的效率,本文对这种多分辨率的方法做了改进并增添额外的约束保证在多层的分割过程中结果的正确性。在视频领域,由于视频中大数据量以及保证结果时空连续性等难点,从视频中提取运动物体是一个很大的挑战。本文将基于多分辨率的图像分割方法扩展到视频领域,减少了传统分割算法的时空耗费。同时利用在低分辨率层次传播中间结果的方法,保证了分割结果的时空连续性。最后,本文提出了基于视频体的多分辨率视频分割系统。它提供了一个有良好交互性的视频编辑用户界面,可以对整个视频体进行切片,标记等操作,并利用基于三维体的分割方法避免了在视频相邻帧间传播结果的不可靠性。实验结果表明,与其他的图像和视频分割算法相比,本文提出的方法大量减少了时间和空间的耗费,并能得到令人满意的分割结果。
[Abstract]:In recent decades, digital image and video processing technology has been widely concerned. Not only professional film and television, animation companies, but also more non-professional users want to edit and process their own media materials. Video segmentation technology plays an important role in material editing. Foreground objects extracted by users from the material can be used in post-processing techniques such as shading, fusion and so on, so as to achieve the purpose of efficient reuse of material. Video segmentation technology and the development of good user interaction material editing tools have great significance. In this paper, image segmentation and video segmentation techniques are studied. The existing image segmentation algorithms are very mature, but with the development of digital high-definition image, video capture equipment, The traditional segmentation technology is difficult to be used in these large amount of data efficiently. Multi-resolution image segmentation method can reduce the number of processing nodes and improve the efficiency of the algorithm. In this paper, the multi-resolution method is improved and additional constraints are added to ensure the correctness of the results in the multi-layer segmentation process. In the field of video, it is a great challenge to extract moving objects from video because of the difficulties of large amount of data in video and the continuity of time and space. In this paper, the method of image segmentation based on multi-resolution is extended to the field of video. The space-time cost of the traditional segmentation algorithm is reduced, and the spatio-temporal continuity of the segmentation results is ensured by using the method of spreading the intermediate results at the low resolution level. Finally, this paper presents a multi-resolution video segmentation system based on video volume, which provides a good interactive video editing user interface, which can slice and mark the whole video body. The method based on 3D volume is used to avoid the unreliability of the results propagating between adjacent frames. The experimental results show that compared with other image and video segmentation algorithms, the proposed method greatly reduces the cost of time and space. Satisfactory segmentation results can be obtained.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：TP391.41

【参考文献】