By counting the detected particles in atomization field holographic image at different threshold levels, a threshold estimation algorithm, which is based on detected particle number varying with the threshold level, is modeled and the rules for optimizing threshold are generalized. For particle field holography with particle size less than 20 μm , the position corresponding to the smallest value in the difference curve of detected particle number varying with threshold levels, is defined as the best threshold. The results from a practical experiment show that this method can work well under such conditions.