图像视频分割方法介绍.ppt

单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,*,问题描述,研究内容,图像和视频中像素点的,聚类,非监督式的学习，共生模型(Generative),图像和视频中像素点的,分类,监督式的学习，判别模型(Discriminative),应用,计算机视觉中的基础性问题,为高层的视觉应用或理解提供支持,图像/视频编辑,多媒体内容检索,物体识别,三维重构,分割方向分类,图像分割,图像分割、过分割、场景解析,图像前景分割,物体分割、交互式分割、前景抠图,视频分割,运动分割、视频场景解析,视频前景分割,视频物体分割、交互式分割、前景抠图,图像共分割,边缘检测,分割方向分类,图像分割,图像分割、过分割、场景解析,图像前景分割,物体分割、交互式分割、前景抠图,视频分割,运动分割、视频场景解析,视频前景分割,视频物体分割、交互式分割、前景抠图,图像共分割,边缘检测,图像分割,场景解析,像素点的分类,监督式，全自动,图像内各个物体的联合识别和分割,方法分类,二阶条件随机场+过分割块,Photoup(Cornell,Siggraph2005,CVPR2006),TextonBoost(Microsoft,ECCV2006),Stair(Stanford,CVPR2008,ICCV2009,NIPS2009),高阶条件随机场,ALE(Oxford,CVPR2008,ECCV2010),过分割块选取(整形规划问题),Stanford,Efficiently Selecting Regions for Scene Understanding.,CVPR:2010.,Maryland,Piecing Together the Segmentation Jigsaw using Context.CVPR,2011,其他：结合场景识别、物体共存性,测试集,MSRC,MIT LabelMe,MIT CBCL Scene Parsing,Stanford Stair,分割方向分类,图像分割,图像分割、过分割、场景解析,图像前景分割,物体分割、交互式分割、前景抠图,视频分割运动分割、视频场景解析,视频前景分割,视频物体分割、交互式分割、前景抠图,图像共分割,边缘检测,物体分割,像素点的分类,监督式、全自动,图像内单个物体的分割,车辆、行人、路牌、花、茶杯等等,分类,轮廓匹配,Accurate Object Localization with Shape Masks.CVPR,2006,Simultaneous Object Detection and Segmentation by Boosting Local Shape Feature based Classifier.CVPR,2007,Adaptive Contour Features in Oriented Granular Space for Human Detection and Segmentation.CVPR,2009,条件随机场+过分割块/像素点+外观特征,Fulkerson B,Vedaldi A,Soatto S:Class Segmentation and Object Localization with Superpixel Neighborhoods.,ICCV:2009.,Aldavert D,Ramisa A,de Mantaras RL,Toledo R:Fast and Robust Object Segmentation with the Integral Linear Classifier.,CVPR:2010.,轮廓匹配结合外观聚类,Combining Top-down and Bottom-up Segmentation.,CVPR:2004,OBJ CUT.CVPR,2005,Object Segmentation by Alignment of Poselet Activations to Image Contours.,CVPR,2011,测试集,VOC PASCAL,Graz Person,Penn Person,Weizmann Horse,Oxford Cow,CalTech Bird,Cambridge Flower.,图像前景分割,图像前景分割,交互式分割,像素点的分类/聚类,监督式？手动,图像内单个物体的分割,方法分类：,马尔科夫随机场下的优化,GraphCut(ICCV2001),GrabCut(Siggraph2004),Lazy Snapping(Siggraph2004),Active Segmentation(CVPR2007),基于谱聚类的方法,Random Walk(PAMI 2009),Biased Normalized Cuts(CVPR2011),基于Level Set的方法,多针对医学图像处理,Morphological Snakes(CVPR2010),Segmentation from a Box(ICCV2011),前景抠图,软分割方法,Alpha通道，0=alpha=2),多个物体(K=2),非监督，全自动,主要方法,基于直方图相似性,Cosegmentation of image pairs by histogram matching.CVPR(2006),Half-Integrality based Algorithms for Cosegmentation of Images.,CVPR,2009,基于轮廓相似性,LOCUS-Learning Object Classes with Unsupervised Segmentation.,ICCV:2005,Detecting and Sketching the Common.,CVPR,2010,基于潜在主题挖掘,Using Multiple Segmentations to Discover Objects and their Extent in Image Collections.,CVPR:2006,Image Segmentation with Topic Random Field.,ECCV,2010,基于区域匹配,A Hybrid Graph Model for Unsupervised Object Segmentation.,ICCV,2007,Unsupervised Detection and Segmentation of Identical Objects.,CVPR:2010,基于突出区域,BiCoS:A Bi-level Co-Segmentation Method for Image Classification.,ICCV,2011,Object Cosegmentation.,CVPR,2011,测试集,MSRC;Caltech Bird;Cambridge Flower;Weizmann Horse;I-Cosegment Dataset;Pairs-segment,分割方向分类,图像分割,图像分割、过分割、场景解析,图像前景分割,物体分割、交互式分割、前景抠图,视频分割,运动分割、视频场景解析,视频前景分割,视频物体分割、交互式分割、前景抠图,图像共分割,边缘检测,边缘检测,边缘检测,应用在基于轮廓的物体识别、匹配、分割,图像分割的对偶问题？,主要方法,早期的:Canny,Sobel等,基于监督学习的方法,Figure/Ground Assignment in Natural Images.,ECCV,2006,Supervised Learning of Edges and Object Boundaries.,CVPR:2006,Efficient,High-Quality Image Contour Detection.,ICCV:2009,测试集：UC Berkeley Image Segmentation,遮挡边缘检测法,Probabilistic Occlusion Boundary Detection on Spatio-Temporal Lattices.,ICCV,2009,Local Occlusion Detection Under Deformations Using Topological Invariants.,ECCV,2010,Occlusion Boundary Detection and Figure/Ground Assignment from Optical Flow.,CVPR,2011,测试集：CMU Occlusion Boundary Dataset,总结,图像/视频分割,标注问题：二值或者多值的最优化求解；,聚类问题：共生模型，包括K-Mean,Plsa,Lda,Ncut(谱聚类)；,分类问题：监督学习，判别模型，包括SVM、Boosting、Random Forest；,满足马尔科夫性：建立MRF/CRF进行全局优化,提高效率：过分割块为计算结点；,特征：颜色、纹理、光流、深度、SIFT、HOG、自相似特征、BOW(Bag of Words)。,迭代求精：,EM,迭代法解决模型参数和分割问题,