Oxygen and nutrient delivery to living tissues, as well as metabolic waste removal, are essentially determined by the dynamics of blood flow in microvascular networks. In these vessels, measuring the velocity distribution of red blood cells (RBCs) is still challenging. One of the most popular techniques used for that purpose is the Dual-Slit (DS), a temporal correlation technique, first introduced by Wayland and Johnson (1967):the vessel under study is trans-illuminated and two photo-sensors (slits) are positioned, separated by a known distance, Ls, along the vessel axis. The time modulation of light is recorded at both positions. A cross-correlation velocity, Vds=Ls/Td, is obtained, where Td is the time delay for which the cross-correlation between the two signals is maximum. However, RBCs are positioned at different depths within the channel and thus move at different velocities. Baker and Wayland (1974) suggested that Vds is related to a dynamic averaged velocity,but this has never been proved. The aim of this work is to determine the relationship between the measured velocity Vds and the actual velocity scales of the flow. For that purpose, the DS technique is first optimized using sequences of synthetic images representing RBCs flow. By this way, all the parameters characterizing the RBCs flow, including the shape of their velocity profile in the direction parallel to the incident light beam, which is inaccessible to the observer in real experiments, are controlled. The DS is then applied to in vitro RBCs flows in microchannels.