Crystal size distribution is important for the production of pharmaceuticals and many other products in a number of industries. A new method to estimate the crystal size distribution is developed based on crystal photomicrographs acquired from an optical microscope imaging system. The calibration of the imaging system is implemented by using a micrometer as a reference. Then particle size information is extracted from each of a set of crystal images through image processing techniques. Finally, according to the least square principle, different distribution functions are adopted to fit the particle size distribution respectively and the result with the minimum square error sum is chosen as the true size distribution. An industrial application shows that the proposed method can provide a satisfactory estimation for the particle size distribution.