Introduction

In this work, we tested and improved an existing algorithm for motion magnification and refined its application in amplifying eye vibrations. 

Experiment

Fourier transformations are a type of mathematical transformation applied to a function which breaks the function down into frequencies present in the function. It is often used in functions for space or time to transform the respective domain into a frequency domain. For functions of space, high-frequency areas will mean sharp edges or fine details, while low-frequency areas will mean smoother areas. 

Using the FFT function built into numpy, a python library, videos can be decomposed intro frequency data maps. Analyzing these and how they change as the video progresses allows for different movements to be identified and examined better and finally amplified to sense vibrations.

In order to handle noise in videos, either Gaussian or Laplacian blurring is used. Gaussian is used when movements are smoother or gradual, as they tend to have lowe frequencies, while Laplacian is used for sharper movements or vibrations, as they tend to have higher frequencies.

Conlusion, Application and Future Outlook

Although there is a relationship between eye strain and eye vibrations, the current dataset is not enough to verify whether or not this method of motion magnification is fully applicable. However, the use of fourier transformations in order to amplify motion has a wide range of applications. It may be used in other similar medical fields, such as early diagnosis of Parkinson's or other neurological diseases. It also holds great potential outside the medical field; being able to accurately see different motion and vibrations normally invisible can have automated, further increasing its effectiveness, alongside refinement of the denoising algorithm and automation of the initial parameters being set.