Frequency vibraimage (FV) characterized the frequency of object vibration [1,2]. It does not mean that FV measures the pixel oscillation frequency, correctly it measures the frequency of signal changing (Hz) in every image pixel for frame difference accumulation period. FV calculates by the following equation:
∆ i = |Ux,y,i – Ux,y(i+1)| - D
where: U x,y,i - the real signal of x,y element in the frame with number i; U x,y (i+1) - the real signal of x,y element in the frame with number i+1; N - the number of frames being processed in the frame sequence; D - the threshold value, typically equal to the image noise level. Fin - video image processing frequency; ∆ i - pixel frame difference.
As AV, real FV calculation algorithm is more complicated than given equation and includes a number of filtered settings and adjusting . FV also depends on several factors, such as object movement, accumulation period (N), object illumination, object optical contrast, etc. But FV dependence from this factors differs from AV, so this two primary types of vibraimage gives more information of object vibration. Vibraimage software displays FV by the standard 256 gradation scale, proportional to video frame input frequency.
Frequency vibraimage samples captured with different accumulated frames number N(10, 50, 100) are below
Together amplitude vibraimage (AV) and frequency vibraimage are the base in vibraimaging processing and head movements characteristics.
References 1. Viktor Minkin. Vibraimage (Russian language). Renome, 2007. 2. US Patent 7346227, Method and device for image transformation. Viktor Minkin, et al. 3. Emotion Recognition System VibraImage Version 7.0 Manual, Elsys Corp., publications, 2008