分割的基本概念

下载 0

乐乐

发布于

3555

人观看

#信息技术

本章节介绍了计算机视觉分割的基本概念，从两个方面来，分割和聚合，介绍了分割和聚合的具体知识，怎么来实施这两步骤，另外关于分割部分，介绍了知觉组织的格式塔理论，关于这一理论的由来以及如何去利用这个理论。

展开查看详情

1 . Lecture 12: Clustering and Segmenta5on Professor Fei-‐Fei Li Stanford Vision Lab Fei-Fei Li! Lecture 12 - 1 ! 28-‐Oct-‐14

2 . What we will learn today •  Introduc5on to segmenta5on and clustering •  Gestalt theory for perceptual grouping •  Agglomera5ve clustering Reading: [FP] Chapters: 14.2, 14.4 Fei-Fei Li! Lecture 12 - 2 ! 28-‐Oct-‐14

3 .Fei-Fei Li! Lecture 12 - 3 ! 28-‐Oct-‐14

4 . Image Segmenta5on •  Goal: iden5fy groups of pixels that go together Slide credit: Steve Seitz, Kristen Grauman Fei-Fei Li! Lecture 12 - 4 ! 28-‐Oct-‐14

5 . The Goals of Segmenta5on •  Separate image into coherent “objects” Image Human segmenta5on Slide credit: Svetlana Lazebnik Fei-Fei Li! Lecture 12 - 5 ! 28-‐Oct-‐14

6 . The Goals of Segmenta5on •  Separate image into coherent “objects” •  Group together similar-‐looking pixels for eﬃciency of further processing Slide credit: Svetlana Lazebnik “superpixels” X. Ren and J. Malik. Learning a classiﬁca5on model for segmenta5on. ICCV 2003. Fei-Fei Li! Lecture 12 - 6 ! 28-‐Oct-‐14

7 . Segmenta5on for feature support 50x50 Patch 50x50 Patch Slide: Derek Hoiem Fei-Fei Li! Lecture 12 - 7 ! 28-‐Oct-‐14

8 . Segmenta5on for eﬃciency [Felzenszwalb and Huttenlocher 2004] [Hoiem et al. 2005, Mori 2005] [Shi and Malik 2001] Slide: Derek Hoiem Fei-Fei Li! Lecture 12 - 8 ! 28-‐Oct-‐14

9 . Segmenta5on as a result Rother et al. 2004 Fei-Fei Li! Lecture 12 - 9 ! 28-‐Oct-‐14

10 . Types of segmenta5ons Oversegmentation Undersegmentation Multiple Segmentations Fei-Fei Li! Lecture 12 - 10 ! 28-‐Oct-‐14

11 . One way to think about “segmenta5on” is Clustering Clustering: group together similar points and represent them with a single token Key Challenges: 1) What makes two points/images/patches similar? 2) How do we compute an overall grouping from pairwise similari5es? Slide: Derek Hoiem Fei-Fei Li! Lecture 12 - 11 ! 28-‐Oct-‐14

12 . Why do we cluster? •  Summarizing data –  Look at large amounts of data –  Patch-‐based compression or denoising –  Represent a large con5nuous vector with the cluster number •  Coun2ng –  Histograms of texture, color, SIFT vectors •  Segmenta2on –  Separate the image into diﬀerent regions •  Predic2on –  Images in the same cluster may have the same labels Slide: Derek Hoiem Fei-Fei Li! Lecture 12 - 12 ! 28-‐Oct-‐14

13 . How do we cluster? •  Agglomera5ve clustering –  Start with each point as its own cluster and itera5vely merge the closest clusters •  K-‐means (next lecture) –  Itera5vely re-‐assign points to the nearest cluster center •  Mean-‐shig clustering (next lecture) –  Es5mate modes of pdf •  Spectral clustering (CS231a, winter quarter) –  Split the nodes in a graph based on assigned links with similarity weights Fei-Fei Li! Lecture 12 - 13 ! 28-‐Oct-‐14

14 . General ideas •  Tokens –  whatever we need to group (pixels, points, surface elements, etc., etc.) •  Bokom up clustering –  tokens belong together because they are locally coherent •  Top down clustering –  tokens belong together because they lie on the same visual en5ty (object, scene…) > These two are not mutually exclusive Fei-Fei Li! Lecture 12 - 14 ! 28-‐Oct-‐14

15 . Examples of Grouping in Vision Grouping video frames into shots Determining image regions What things should Figure-‐ground be grouped? What cues indicate groups? Slide credit: Kristen Grauman Object-‐level grouping Fei-Fei Li! Lecture 12 - 15 ! 28-‐Oct-‐14

16 . Similarity Slide credit: Kristen Grauman Fei-Fei Li! Lecture 12 - 16 ! 28-‐Oct-‐14

17 . Symmetry Slide credit: Kristen Grauman Fei-Fei Li! Lecture 12 - 17 ! 28-‐Oct-‐14

18 . Common Fate Image credit: Arthus-‐Bertrand (via F. Durand) Slide credit: Kristen Grauman Fei-Fei Li! Lecture 12 - 18 ! 28-‐Oct-‐14

19 . Proximity Slide credit: Kristen Grauman Fei-Fei Li! Lecture 12 - 19 ! 28-‐Oct-‐14

20 . Muller-‐Lyer Illusion •  What makes the bokom line look longer than the top line? Fei-Fei Li! Lecture 12 - 20 ! 28-‐Oct-‐14

21 . What we will learn today •  Introduc5on to segmenta5on and clustering •  Gestalt theory for perceptual grouping •  Agglomera5ve clustering Fei-Fei Li! Lecture 12 - 21 ! 28-‐Oct-‐14

22 . The Gestalt School •  Grouping is key to visual percep5on •  Elements in a collec5on can have proper5es that result from rela5onships –  “The whole is greater than the sum of its parts” Illusory/subjec5ve Occlusion contours Slide credit: Svetlana Lazebnik Familiar conﬁgura5on hkp://en.wikipedia.org/wiki/Gestalt_psychology Fei-Fei Li! Lecture 12 - 22 ! 28-‐Oct-‐14

23 . Gestalt Theory •  Gestalt: whole or group –  Whole is greater than sum of its parts –  Rela5onships among parts can yield new proper5es/features •  Psychologists iden5ﬁed series of factors that predispose set of elements to be grouped (by human visual system) “I stand at the window and see a house, trees, sky. Theoretically I might say there were 327 brightnesses and nuances of colour. Do I have "327"? No. I have sky, house, and trees.” Max Wertheimer (1880-1943) Untersuchungen zur Lehre von der Gestalt, Psychologische Forschung, Vol. 4, pp. 301-350, 1923 http://psy.ed.asu.edu/~classics/Wertheimer/Forms/forms.htm Fei-Fei Li! Lecture 12 - 23 ! 28-‐Oct-‐14

24 . Gestalt Factors Image source: Forsyth & Ponce •  These factors make intui5ve sense, but are very diﬃcult to translate into algorithms. Fei-Fei Li! Lecture 12 - 24 ! 28-‐Oct-‐14

25 . Con5nuity through Occlusion Cues Fei-Fei Li! Lecture 12 - 25 ! 28-‐Oct-‐14

26 . Con5nuity through Occlusion Cues Con5nuity, explana5on by occlusion Fei-Fei Li! Lecture 12 - 26 ! 28-‐Oct-‐14

27 . Con5nuity through Occlusion Cues Image source: Forsyth & Ponce Fei-Fei Li! Lecture 12 - 27 ! 28-‐Oct-‐14

28 . Con5nuity through Occlusion Cues Image source: Forsyth & Ponce Fei-Fei Li! Lecture 12 - 28 ! 28-‐Oct-‐14

29 . Figure-‐Ground Discrimina5on Fei-Fei Li! Lecture 12 - 29 ! 28-‐Oct-‐14

6点赞

2收藏

0下载