VidMag

 

Aims

 

VidMag is an implementation of the paper "Eulerian Video Magnification for Revealing Subtle Changes in the World" by Wu et al. It enables users to visualize and analyze subtle changes in color and/or motion in videos by magnifying them.

 

In contrast to the commandline-based sample implementation in Matlab by the authors of the paper, we decided to implement the Eulerian video magnification using C++ with OpenCV and added an intutive user interface with Qt for easier usage. The processing of a given video may take a few seconds due to the nature of the algorithm, but since we use C++ and OpenCV the processing is much faster than with Matlab.

 

 

Interface & Usage

 

 

The GUI offers the usual basic interactions for handling and analyzing videos, such as loading a video from the filesystem ("Load" in the lower left corner), playing a video ("Play"), loading parameter-presets ("Load" in the upper right corner), and choosing different processing parameters as well as the type of Eulerian video magnification. Furthermore, it is possible to create a mask to analyze only certain areas of interest.

 

For using our implementation, one has to initially load a videofile via the load-button in the lower left corner. After this, a parameter-preset may be loaded or the processing parameters are selected manually. For beginners we advice to start with a paramter-present and vary the paramters as desired. Once the parameters are all set, a mask may be applied. The video can be processed via the process button in the lower right corner. The processed video is automatically saved to ./Data and can be played via the play button. Different characteristics of the video may be explored by changing the parameters and processing the video again.

 

 

Development and Libraries

 

Our implementation was developed in Visual Studio 2013 with OpenCV 3.1 for Windows x86. For the GUI we used Qt 5.6 and the corresponding Visual Studio Addin. The implementation was tested on Windows 7 x64 and Windows 10 x64.

 

 

Code and Documentation

 

Our executables and some of the testvideos by the original authors can be found here (make sure to keep the order of folders intact as the processed videos are saved to the Data-folder). The corresponding sourcecode can be downloaded here. For further information about implementation details we refer to the code documentation.