A new method of particle image detection has been implemented and tested for the use in the particle tracking velocimetry. The method benefits from a new algorithm of extracting individual particle images from a massive concentration of particles recorded in the PIV recordings. With this algorithm, the binary threshold level is determined particle by particle in such a way that the mean gray level of every particle is kept at a constant degree of contrast against the background. So the particles of any size, shape or gray levels are extracted. In addition, the algorithm is applicable to a variety of particle images, including those suffering from background noise. In the latter half of the paper, the new algorithm is further improved in order to speed up the computation and facilitate the estimation of the preset parameters.