Visualization of the smallest blood vessels in the brain, capillaries, and assessment of the blood flow rate in them is important in many physiological studies. However, it is in this case that conventional label-free imaging methods fail since both the number and velocity of red blood cells in the capillaries are often too low. We present a label-free method of capillary blood flow analysis aimed at detecting and counting each single red blood cell in order to build a very detailed map of the vasculature. Such a map, in turn, enables us to more effectively apply the Particle Image Velocimetry method and make label-free blood flow velocity measurements in the smallest capillaries. Technically, our method is based on the adaptive spatial filtering of each frame of the acquired series of images using adaptive Niblack filtration. As a result of frame-by-frame filtering, we can differentiate single moving RBCs from static image artifacts having a similar size and brightness. We show the method applicability using two different biological models, specifically, the chicken embryo and the mouse brain.