A new image processing method based on image calibration has been developed to visualize clusters in a Plexiglas rectangular circulating fluidized bed (CFB) riser at the fully developed region. A high speed video camera is adopted with FCC particles of 67 mu m used as the bed material. In order to make sure day-to-day shooting images reliable and comparable under the same light illumination during the whole shooting period, a reference plate is made first to verify the light illumination consistency. After the verification, the relationship between solids holdup and image grayscale is calibrated in the same riser with both calibration curve and equation obtained. With given solids holdup thresholds increasing from 0.01 to 0.1 cluster size clearly decreases while cluster solids holdup increases by transforming gray images into binary ones. The variation of cluster profiles with different thresholds also reveals that there is a solids holdup gradient inside the cluster with a lower value at the surface and a higher one in the core. The change in dense phase holdup or cluster population with operating conditions is indicated by the variation of the black area under given solids holdup thresholds, which compare well with the well-known fact reported by many previous studies: higher G(s) and lower U-g are beneficial for clusters with higher solids holdup to form, while lower G, and higher Ug induce clusters to breakup. This consistency proves that the image processing method developed in the present study is effective and very useful. (C) 2014 Elsevier B.V. All rights reserved.