Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Overview of Image Retrieval Hui-Ying Wang. 2/42 Reference Smeulders, A. W., Worring, M., Santini, S., Gupta, A.,, and Jain, R. 2000. “Content-based.

Similar presentations


Presentation on theme: "1 Overview of Image Retrieval Hui-Ying Wang. 2/42 Reference Smeulders, A. W., Worring, M., Santini, S., Gupta, A.,, and Jain, R. 2000. “Content-based."— Presentation transcript:

1 1 Overview of Image Retrieval Hui-Ying Wang

2 2/42 Reference Smeulders, A. W., Worring, M., Santini, S., Gupta, A.,, and Jain, R. 2000. “Content-based image retrieval at the end of the early years.” IEEE Trans. Pattern Analysis and Machine Intelligence 22, 12, 1349–1380. R. Datta, D. Joshi, J. Li and J. Z. Wang, ”Image Retrieval: Ideas, Influences, and Trends of the New Age,” ACM Computing Surveys, 2008, to appear. CVPR 2007 short course: Recognizing and Learning Object Categories http://people.csail.mit.edu/torralba/shortCourseRLOC/inde x.html

3 3/42 Outline Motive Academic and real world Difficulties and problem model Evaluation metrics

4 4/42 Outline Motive Academic and real world Difficulties and problem model Evaluation metrics

5 5/42 Motive Popular electronic device –Digital camera By-product –Digital photos Need –Organization Key: filenames? dates?

6 6/42 Outline Motive Academic and real world Difficulties and problem model Evaluation metrics

7 7/42 Publications in CBIR

8 8/42 Real-world system Search engines –Google Image (2.2 b) –Picsearch (1.7 b) –Yahoo! Images (1.6 b) –AltaVista –Ask Images Online albums –Flickr –Riya –Webshots Shopping –like

9 9/42 Real-world system Search engines –Google Image (2.2 b) –Picsearch (1.7 b) –Yahoo! Images (1.6 b) –AltaVista –Ask Images Online albums –Flickr –Riya –Webshots Shopping –like

10 10/42 Google Images

11 11/42 Google Image Labeler

12 12/42 Picsearch

13 13/42 Yahoo! Images

14 14/42 AltaVista

15 15/42 Ask Images

16 16/42 Real-world system Search engines –Google Image (2.2 b) –Picsearch (1.7 b) –Yahoo! Images (1.6 b) –AltaVista –Ask Images Online albums –Flickr –Riya –Webshots Shopping –like

17 17/42 Flickr

18 18/42 Webshots

19 19/42 Riya

20 20/42 Real-world system Search engines –Google Image (2.2 b) –Picsearch (1.7 b) –Yahoo! Images (1.6 b) –AltaVista –Ask Images Online albums –Flickr –Riya –Webshots Shopping –like

21 21/42 like

22 22/42 Outline Motive Academic and real world Difficulties and problem model Evaluation metrics

23 23/42 Challenges view point variation scale illumination deformation occlusion

24 24/42 Goal real object sensory gap digital record interpretation semantic gap extraction human vision computer vision

25 25/42 Core problems How to describe an image How to assess the similarity

26 26/42 Some features Global features –MPEG-7 Color Layout Descriptor Edge Histogram Descriptor Homogeneous Texture Descriptor Summarizing local features –Bag of Features

27 27/42 Some features Global features –MPEG-7 Color Layout Descriptor Edge Histogram Descriptor Homogeneous Texture Descriptor Summarizing local features –Bag of Features

28 28/42 Color Layout Descriptor - Presentation MPEG-7

29 29/42 Color Layout Descriptor - Similarity

30 30/42 Some features Global features –MPEG-7 Color Layout Descriptor Edge Histogram Descriptor Homogeneous Texture Descriptor Summarizing local features –Bag of Features

31 31/42 Edge Histogram Descriptor - Presentation MPEG-7

32 32/42 Edge Histogram Descriptor - Similarity

33 33/42 Some features Global features –MPEG-7 Color Layout Descriptor Edge Histogram Descriptor Homogeneous Texture Descriptor Summarizing local features –Bag of Features

34 34/42 Homogeneous Texture Descriptor - Presentation e: log-scaled sum of the squares of Gabor-filtered Fourier transform coefficients d: log-scaled standard deviation of the squares of Gabor-filtered Fourier transform coefficients Human Vision System Fourier transform Gabor function f DC : mean deviation f SD : standard deviation

35 35/42 Homogeneous Texture Descriptor - Similarity

36 36/42 Some features Global features –MPEG-7 Color Layout Descriptor Edge Histogram Descriptor Homogeneous Texture Descriptor Summarizing local features –Bag of Features

37 37/42 Local feature Detected keypoints –spatial relationship fully independent (ex: bag of features) fully connected

38 38/42 Bag of Features

39 39/42 Outline Motive Academic and real world Difficulties and problem model Evaluation metrics

40 40/42 Evaluation (1/2) Standard –Precision # of retrieved positive images / # of total retrieved images –Recall # of retrieved positive images / # of total positive images

41 41/42 Evaluation (1/2) When number of retrieved images increase –Recall ↑ Precision ↓ Average precision (AP) –The area under the precision-recall curve for a query precision recall 1 1 AP

42 42 The end ~ Thank you


Download ppt "1 Overview of Image Retrieval Hui-Ying Wang. 2/42 Reference Smeulders, A. W., Worring, M., Santini, S., Gupta, A.,, and Jain, R. 2000. “Content-based."

Similar presentations


Ads by Google