This paper describes limit possibilities of modern cooled thermal imaging cameras as a tool for estimation of blood flow oscillations at the surface of living body. Skin temperature oscillations, as we assumed, are a consequence of the blood flow oscillations. We considered the temperature sensitivity 0.01-0.02 °C as a typical for the most of modern cooled long wave thermal imaging cameras. Fourier filter used to investigate the temperature signal separately within endothelial, neurogenic, myogenic, respiratory and cardiac frequency ranges. The level of temporal noise has been estimated during measurements of no living body with stabilized temperature ~ 24°C. The level of temperature oscillations has been calculated for the group of healthy subjects within each frequency range. Thus, we were able to determine signal-to-noise ratio within frequency band [0.001, 1] Hz. As a result, we determine that skin temperature oscillations measured by thermal imaging camera with sensitivity 0.02°C have the upper frequency limit ~ 0.2 Hz. In other words, within the respiratory and cardiac frequency ranges of blood flow oscillations the noise level exceeds signal one, and temperature measurements at the skin surface are practically useless. The endothelial, neurogenic and myogenic components of the temperature oscillations contain ~98% of the total spectral power of the signal. We have plot the empirical extrapolated curve of sensitivity of thermal imaging camera vs. frequency of the temperature oscillations. The data analysis shows that measurements of skin temperature oscillations within respiratory and cardiac ranges require the temperature sensitivity at least ~ 0.01°C and 0.001°C, respectively.