This paper presents Eulerian video magnification, a method to reveal subtle changes in videos that are imperceptible to the human eye. The method involves spatial decomposition and temporal filtering of video frames to amplify hidden information. By applying a bandpass filter to temporal frequency bands of interest, the method can magnify small motions and color variations, such as blood flow in the face or subtle movements in a baby's chest. The approach does not require feature tracking or motion estimation, making it efficient and suitable for real-time applications. The method is based on a linear approximation of the brightness constancy assumption used in optical flow algorithms. It can be applied to various applications, including medical monitoring, where subtle changes in facial blood flow can indicate health issues. The method is robust and can handle different spatial and temporal frequencies, making it suitable for both small and large motions. The paper also compares Eulerian and Lagrangian motion magnification approaches, showing that Eulerian methods are more effective for small amplifications and larger noise levels. The method has been tested on various videos, including those of faces, babies, and musical instruments, demonstrating its ability to reveal subtle changes in motion and color. The results show that the method can amplify subtle motions and color changes, making them visible to the human eye. The paper also discusses the limitations of the method, including the need for careful selection of temporal filters and amplification factors to avoid artifacts. The method is implemented in C++ and can process videos at 45 frames per second on a standard laptop. The code and videos are available for further study.This paper presents Eulerian video magnification, a method to reveal subtle changes in videos that are imperceptible to the human eye. The method involves spatial decomposition and temporal filtering of video frames to amplify hidden information. By applying a bandpass filter to temporal frequency bands of interest, the method can magnify small motions and color variations, such as blood flow in the face or subtle movements in a baby's chest. The approach does not require feature tracking or motion estimation, making it efficient and suitable for real-time applications. The method is based on a linear approximation of the brightness constancy assumption used in optical flow algorithms. It can be applied to various applications, including medical monitoring, where subtle changes in facial blood flow can indicate health issues. The method is robust and can handle different spatial and temporal frequencies, making it suitable for both small and large motions. The paper also compares Eulerian and Lagrangian motion magnification approaches, showing that Eulerian methods are more effective for small amplifications and larger noise levels. The method has been tested on various videos, including those of faces, babies, and musical instruments, demonstrating its ability to reveal subtle changes in motion and color. The results show that the method can amplify subtle motions and color changes, making them visible to the human eye. The paper also discusses the limitations of the method, including the need for careful selection of temporal filters and amplification factors to avoid artifacts. The method is implemented in C++ and can process videos at 45 frames per second on a standard laptop. The code and videos are available for further study.