Skip to content

Topic 10: Video magnification

Getting comfortable

  1. Browse this website on video magnification

  2. Read this paper [2] and download its Matlab code and the corresponding video data which are available here.

Assignments part one

  1. Analyze whats going on spatially in [2]. In the Matlab code and in the paper [2] there are several image pyramids used. For each pyramid: 1) Analyze its properties; 2) motivate the reasons why it is used and 3) Compare its properties to others (i.e.: advantages/disadvantages). (The lectures on image pyramids may prove useful here)

  2. Analyze whats going on temporally in [2]. In the Matlab code and in the paper [2] there are several temporal filters used. For each temporal filter: 1) Analyze its properties; 2) motivate the reasons why it is used and 3) Compare its properties to others (i.e.: advantages/disadvantages).

  3. Briefly describe the difference between Eulerian and Lagrangian motion. For each method: 1) Analyze its properties and 2) Compare its properties to the other (i.e.: advantages/disadvantages).

  4. Reproduce the results of [2]. Use the provided software to validate that you can reproduce the results from the paper [2]. Now, make two video recordings yourself: a) a color-magnification video and b) a motion-magnification video. Demonstrate (how?) and analyze (how?) what the results are on those videos a and b. Make sure to also show the original videos and to motivate parameter settings.

The next phase

  1. Browse the following site and watch the “supplemental video” here.

  2. Read this follow-up paper [1].

Assignments part two

  1. Analyze whats going on spatially in [1]. In [1] there is a different image pyramid used than in [2]. For the pyramid in [1]: 1) Analyze and describe its properties; 2) motivate the reasons why it is used and 3) Compare its properties to the pyramids used in [2] (i.e.: advantages/disadvantages).

  2. Implement [1]. Make your own version of the method of [1], where you may use existing image pyramid implementations. Note that we know that the code of [1] is available on request, and there are some implementations available on the internet.

  3. Demonstrate the results of [1]. Use your software on the two video recordings that you made yourself: a) a color-magnification video and b) a motion-magnification video. Demonstrate (how?) and analyze (how?) what the results are on those videos a and b. Make sure to also show the original videos and to motivate parameter settings and compare against your results of [2] in assignment 4.

References

[1] NWadhwa, M Rubinstein, F Durand, andWT Freeman. Phase-based video motion processing. ACM-TOG, 32(4):80, 2013.

[2] Hao-Yu Wu, Michael Rubinstein, Eugene Shih, John Guttag, Fredo Durand, and William T. Freeman. Eulerian video magnification for revealing subtle changes in the world. ACM Trans. Graph. (Proceedings SIGGRAPH 2012), 31(4), 2012.

Last update: 2023-04-17