In this paper, a model-based reconstruction technique is proposed to simultaneously measure the relative deoxyhemoglobin concentration and the relative blood flow velocity in cerebral cortex. With the help of this model-based reconstruction technique, artifacts due to nonuniform laser illumination and curvature of cortex are efficiently corrected. The results of relative deoxyhemoglobin concentration and relative blood flow velocity are then used to detect and distinguish cerebral arteries and veins. In an experimental study on rat, cerebral blood vessels are segmented from the reconstructed blood flow image by Otsu multiple threshold method. Afterwards, arteries and veins are distinguished by a simple fuzzy criterion based on the information of relative deoxyhemoglobin concentration.