Abstract: A constrained regularization method and the CONTIN algorithm that based on it are presented for inversion computation of particle size distributions in photon correlation spectroscopy. CONTIN algorithm has been tested with simulated data corresponding to unimodal and bimodal size distributions. We study the spread effect of CONTIN, the relationship between the effect and particle size. The algorithm has also been used in experimental analysis. CONTIN's results are compared with those of Exponential Sampling method and Nonnegative Constrained Least Squares method (NNLS). Our study shows that CONTIN's resolution power is related to particle size and is poor compared with NNLS. Choosing a small regularization parameter or shortening the inversion range properly can help improve the inversion quality.