Figure 2.

Comparison of the color magnification results on face sequence in presence of a nonlinear motion. (a) Top: An original frame from the input video. Bottom: A spatio-temporal $X-T$ slice of the video along the red line marked on the input frame. (b) The linear EVM [6] intensity processing result and the corresponding spatio-temporal slice. (c) Color magnification using video acceleration magnification [17] and the $X-T$ slice along the red line. (d) Our result of video color magnification and the temporal variation at the location indicated by the red stripe. Note that our algorithm is better able to magnify the intensity for video with a non-linear motion.