We propose a full-field optical method for the label-free and quantitative mapping of the velocities of red blood cells (RBCs) in capillaries. It integrates spatiotemporal demodulation and an autocorrelation algorithm, and measures RBC velocity according to the ratio of RBC length to lag time. Conventionally, RBC length is assumed to be a constant and lag time is taken as a variable, while our method treats both of them as variables. We use temporal demodulation and the Butterworth spatial filter to separate RBC signal from background signal, based on which we obtain the RBC length by image segmentation and lag time by autocorrelation analysis. The RBC velocity calculated now is more accurate. The validity of our method is verified by an in vivo experiment on a mouse ear. Owing to its higher image signal-to-noise ratio, our method can be used for mapping RBC velocity in the turbid tissue case.