Magnified Movements - Using Eulerian Video Magnification with Cultural Heritage Objects

INTRODUCTION

The magnification studies conducted in our research use the MATLAB implementation of the Eulerian Video Magnification (EVM) SIGGRAPH 2012 code developed by the Massachusetts Institute of Technology (MIT). The MATLAB source code is freely available for noncommercial purposes can be downloaded from the MIT SIGGRAPH 2012 website: http://people.csail.mit.edu/mrub/evm/#code. This website provides a lot of useful reference material on the original magnification tests that were conducted at MIT.

This user manual outlines the technical workflow used for examining the effects of changes in relative humidity on cultural heritage materials. This includes filming parameters, editing workflow, and MATLAB software implementation. While initial technical support was provided by the IT administrator at the School of Information at the University of Texas at Austin, the primary workflow required little technical expertise and was carried out by graduate student Ayshea Khan and faculty member Karen Pavelka.

FILMING PARAMETERS

At the time of capture, it is recommended that subjects be recorded at the highest resolution settings possible. Cameras equipped with full-frame sensors will reveal optimal movement after video processing. Mount your camera to a tripod to avoid extraneous motion and ensure that your subject is taking up most of the space in the frame.
While professional grade cameras are recommended for the best results, subtle material movements can still be revealed using more affordable equipment. Our studies were conducted using two consumer grade cameras: A Sony Handycam HDR video camcorder and a Nikon DSLR. We used the Sony camcorder for the initial videos, and later, moved to the Nikon, paired with a Phottix intervalometer, to create higher definition time lapses. While not necessary, time lapsed studies allowed for a longer observation period while minimizing both file size and processing time. All of our videos that were processed through the EVM software were filmed at 30 frames per second at a resolution of 1920 by 1080 pixels.

EDITING WORKFLOW

It is not required that video studies be exported through a video editing software prior to magnification processing. However, since our later studies were taken in the form of time lapses, the photographic sequences needed to be arranged and exported. This was done using iMovie, a freely available video editing application for OS users.

MATLAB IMPLEMENTATION OF SIGGRAPH 2012

All Eulerian Video Magnification processing was performed on a late 2014 iMac with a 4GHz Intel Core i7 processor and 32G 1600 MHz DDR3 memory, running OSX Yosemite 10.10.5. The speed of this machine allowed for a video one minute in length to be processed in roughly forty minutes. While the code allows for bundled video processing, we processed videos individually to avoid computer freezing or failure.

The SIGGRAPH2012 Eulerian Video Magnification code is openly available through MIT (http://people.csail.mit.edu/mrub/evm/#code). Our testing utilized a MATLAB implementation, provided by The University of Texas at Austin. Installation requires limited technical proficiency, as a detailed README text file is include in the code bundle. Graduate student Ayshea Khan was able to navigate to the EVM software through MATLAB with little difficulty.
If MATLAB is not available at your institution, MIT has also provided 64-bit Windows, 64-but Linux, and 64-bit Mac compiled codes that facilitate magnification processing to be performed at the command line without the MATLAB application. These executables also contain user-friendly README text files. However, this instance of the code may require a user with familiarity with command line operation.

Basic Steps for MATLAB EVM Implementation:

  1. Start up MATLAB and change directory to the location of this code.
  2. Run “setpath.m” to add all directory folders to code path.
  3. Add desired videos to Data folder for processing
  4. Run “make.m” to build pyramid toolbox libraries (this is REQUIRED if using Mac OS and MATLAB newer than 2011b).
  5. Run “install.m”
  6. Open “reproduceResults.m” code and adjust desired parameters, using the % to comment out unwanted code.
  7. Parameters are defined in parentheses as (Amplification, Cut Off Wavelength, Frequency1, Frequency2, Frame Rate, Chrominence Scaling)
  8. Run “reproduceResults.m” to process video
  9. All processed video can be found in the 2012Results Folder.