Presentation is loading. Please wait.

Presentation is loading. Please wait.

Image Processing on Line A new way to publish? A new way to organize research in a lab? Jérôme Darbon (UCLA and ENS Cachan), Nicolas Limare (ENS Cachan),

Similar presentations


Presentation on theme: "Image Processing on Line A new way to publish? A new way to organize research in a lab? Jérôme Darbon (UCLA and ENS Cachan), Nicolas Limare (ENS Cachan),"— Presentation transcript:

1 Image Processing on Line A new way to publish? A new way to organize research in a lab? Jérôme Darbon (UCLA and ENS Cachan), Nicolas Limare (ENS Cachan), Jean-Michel Morel (ENS Cachan) 01/10/20091 Image Processing on Line, Minneapolis, ONR meeting 2009 Research partially funded by ONR Grant N00014-97-1-0839

2 Image Processing on Line: main goal 01/10/20092 Image Processing on Line, Minneapolis, ONR meeting 2009 Achieve « reproducible research » and therefore: Allow EVERYBODY to try the algorithms on their own images (including deciders who DO NOT program) Make all results of trials accessible in a Web archive (experiment sharing) Downloadable code Give a list of examples AND counterexamples, and explain them Describe the algorithms carefully (pseudocode+comments) Testing independent of any platform (no code download, no system requirement…) This is particularly adapted to image processing!

3 Means of the project 01/10/20093 Image Processing on Line, Minneapolis, ONR meeting 2009 Online servers, multicore Execution in real/interactive time (less than 20 seconds) Research team on online web tools Some (easy) parallel computing Reorganization of the work flow in a team

4 Reorganization of the work flow in a research team 01/10/20094 Image Processing on Line, Minneapolis, ONR meeting 2009 Math theory and first algorithm Elimination of parameters, private tests Internal web version tested on line by the whole team, feedback on code and theory List of examples and counterexamples established Writing of the web site, green light to online demo Submission of paper, clear ideas on future work

5 01/10/2009 Image Processing on Line, Minneapolis, ONR meeting 2009 5 Lab. Web server Internal lab workshops, testing in group (private) Easy building blocks for assembling a complex chain (video processing stereo…) On line algorithms to contractors (private) Online publication (even for « well known « algorithms Teaching Four functions for a Web server

6 01/10/2009 Image Processing on Line, Minneapolis, ONR meeting 2009 6 Algorithm website Link to code download, C code, several platforms Link to on line demo Link to publication reference or preprint Link to archive of on line trials For each algorithm: a fourfold publication

7 01/10/20097 Image Processing on Line, Minneapolis, ONR meeting 2009 Current state of the site PUBLIC ONLINE (Website and demos): http://mw.cmla.ens-cachan.fr/megawave/algo/:http://mw.cmla.ens-cachan.fr/megawave/algo/ ASIFT : Affine Invariant Image Matching LSD : Line Segment Detection Random Phase Noise : Microtexture Synthesis Cartoon+Texture Image Decomposition Retinex PDE Color Balance INTERNAL WORKSHOPS: Color contrast enhancement by PDE (three algorithms compared) http://mw.cmla.ens-cachan.fr/megawave/demo/retinex_workshop/ NL-means (on test) http://mw.cmla.ens-cachan.fr/megawave/demo/nlmeans/

8 01/10/20098 Image Processing on Line, Minneapolis, ONR meeting 2009 ASIFT: affine invariant image comparison

9 01/10/20099 Image Processing on Line, Minneapolis, ONR meeting 2009 Result of ASIFT:Compared with SIFT:

10 01/10/200910 Image Processing on Line, Minneapolis, ONR meeting 2009 Algorithm Website

11 01/10/200911 Image Processing on Line, Minneapolis, ONR meeting 2009 The website of each algorithms shows and explains the failure cases. For instance for SIFT and ASIFT, failure comparing objects with night and day illumination

12 01/10/200912 Image Processing on Line, Minneapolis, ONR meeting 2009 The online demo gives also access to the online archive. More than 1550 different images have been so far tried by on line users. They are grouped in pages of 50. Here are three examples tried by users, on a simple box, a building and a landscape.

13 01/10/200913 Image Processing on Line, Minneapolis, ONR meeting 2009 By a simple click a closer view of each experiment is available: ASIFT result :SIFT result :

14 01/10/200914 Image Processing on Line, Minneapolis, ONR meeting 2009

15 01/10/200915 Image Processing on Line, Minneapolis, ONR meeting 2009 Line segment detector (LSD), no parameter. More than 850 images in the archive

16 01/10/200916 Image Processing on Line, Minneapolis, ONR meeting 2009 Some results

17 01/10/200917 Image Processing on Line, Minneapolis, ONR meeting 2009 LSD: Some results in the online archive

18 01/10/200918 Image Processing on Line, Minneapolis, ONR meeting 2009

19 01/10/200919 Image Processing on Line, Minneapolis, ONR meeting 2009

20 01/10/200920 Image Processing on Line, Minneapolis, ONR meeting 2009

21 01/10/200921 Image Processing on Line, Minneapolis, ONR meeting 2009

22 01/10/200922 Image Processing on Line, Minneapolis, ONR meeting 2009

23 01/10/200923 Image Processing on Line, Minneapolis, ONR meeting 2009

24 01/10/200924 Image Processing on Line, Minneapolis, ONR meeting 2009

25 01/10/200925 Image Processing on Line, Minneapolis, ONR meeting 2009

26 01/10/200926 Image Processing on Line, Minneapolis, ONR meeting 2009

27 01/10/200927 Image Processing on Line, Minneapolis, ONR meeting 2009 Plans -Make a substantial database of algorithms. New ones. But many classic ones -Pass from atoms to molecules :. Stereo (calibration, stereorectification, matching, reconstruction. Image processing (noise, blur, color, contrast: requires reliable estimates) -Official publication?. Needs a scientific AND technical committee. Needs a union of labs to start (right now three labs volunteer). Needs to fix very carefully the rules for code description, examples, etc.. Needs €€€€€ + $$$$$ VISIT US: http://mw.cmla.ens-cachan.fr/megawave/algo/:http://mw.cmla.ens-cachan.fr/megawave/algo/

28 01/10/200928 Image Processing on Line, Minneapolis, ONR meeting 2009

29 01/10/200929 Image Processing on Line, Minneapolis, ONR meeting 2009

30 01/10/200930 Image Processing on Line, Minneapolis, ONR meeting 2009

31 01/10/200931 Image Processing on Line, Minneapolis, ONR meeting 2009

32 01/10/200932 Image Processing on Line, Minneapolis, ONR meeting 2009

33 01/10/200933 Image Processing on Line, Minneapolis, ONR meeting 2009

34 01/10/200934 Image Processing on Line, Minneapolis, ONR meeting 2009

35 01/10/200935 Image Processing on Line, Minneapolis, ONR meeting 2009

36 01/10/200936 Image Processing on Line, Minneapolis, ONR meeting 2009

37 01/10/200937 Image Processing on Line, Minneapolis, ONR meeting 2009

38 01/10/200938 Image Processing on Line, Minneapolis, ONR meeting 2009

39 01/10/200939 Image Processing on Line, Minneapolis, ONR meeting 2009

40 01/10/200940 Image Processing on Line, Minneapolis, ONR meeting 2009

41 01/10/200941 Image Processing on Line, Minneapolis, ONR meeting 2009

42 01/10/200942 Image Processing on Line, Minneapolis, ONR meeting 2009

43 01/10/200943 Image Processing on Line, Minneapolis, ONR meeting 2009

44 01/10/200944 Image Processing on Line, Minneapolis, ONR meeting 2009

45 01/10/200945 Image Processing on Line, Minneapolis, ONR meeting 2009

46 01/10/200946 Image Processing on Line, Minneapolis, ONR meeting 2009


Download ppt "Image Processing on Line A new way to publish? A new way to organize research in a lab? Jérôme Darbon (UCLA and ENS Cachan), Nicolas Limare (ENS Cachan),"

Similar presentations


Ads by Google