Fluidized beds are used in a wide number of applications, including power plant boilers and chemical facilities. This study proposes a novel particle tracking velocimetry algorithm for semi-dilute suspensions present in fluidized beds. The proposed algorithm is based on thresholding and profile matching algorithms. Image intensity thresholding is used to find regions which need additional image processing. These regions are then processed using interference-based profile matching algorithm to refine the solution quality. The key idea is to limit heavy profile matching computations only to identified clusters to save as much computation time as possible. The method was tested with simulated data and experimental data. The simulations showed that the proposed method was better than the previous boundary arc detection-based method in noisy conditions and cluster sizes ranging from 2 to 6 particles. The difference between the previous approach and the proposed method was small in cluster sizes larger than six particles, even though the proposed method was still slightly better. In low noise conditions with only two particles, the boundary arc detection could outperform the proposed method, but the difference was small. The method was also tested with experimental data from a small cold model fluidized bed. The velocity distributions obtained from the bed are shown for qualitative evaluation. The velocity distributions were realistic, which suggests the usability of the method. The proposed method was also compared to particle image velocimetry to see if both methods produce similar results. The results were divided into histogram intervals according to image intensity that was proportional to local solids volume fraction. The comparison showed that both methods produced similar results, in particular in the low-intensity range, which supports the ability of the method to produce realistic results also in the semi-dilute range where particles form small clusters.