Septic shock induces organ dysfunction by microcirculatory disturbance. Observation and quantification of microcirculation are expected to be effective for the diagnosis of septic shock. Sidestream dark-filed (SDF) imaging is a suitable technique for observation of microcirculation. It can noninvasively visualize red blood cells (RBCs) of microcirculation. We are developing early diagnostic criteria for septic shock from microcirculation SDF images. As an initial study, we use the blood flow velocity estimated from the images as a diagnostic criteria. However, low contrast quality and subject's movement disturb the blood flow velocity estimation. In this paper, we present a procedure of image processing for a stable estimation of the blood flow velocity. In the procedure, we first perform a robust principal component analysis (RPCA) as a preprocessing. RPCA decomposes a motion picture into a low-rank (L) component and a sparse (S) component. The S component images clearly expresses RBCs flow and is used for the velocity estimation. The temporal change of the intensity profile along the vessel was analyzed by Hough transform to estimate the blood flow velocity is. The proposed procedure was examined with dorsal microcirculation of septic model rats and a sham rat. As a result, the decrease in blood flow velocity of the septic rats after 17 hours was greater than that of the sham. It was also suggested that blood flow velocity might be faster index of septic shock reaction earlier than lactic acid value. These results suggest that the velocity estimation is reasonable for diagnosis of septic shock.