Bubble size distributions in bubble‐liquid reactors are essentially determined by the multiphase turbulent hydrodynamic behaviors, which take greatly effects on the heat and mass transfer. An improved image procession algorithm, effectively combined for the first time, color space transformation, shadow detection and removal, and watershed segmentation to handle the raw pictures captured by high‐speed camera was proposed to characterize the distributions of bubble sizes. The hue–saturation–intensity color space approach to detect bubble edges and shadow regions for their intensity was set to as input value for processing background subtraction. Furthermore, the removal of bubble shadow was successfully performed by means of extracting shadow regions to attain the satisfied compensation images, a watershed transformation function was also developed to segment the overlapping bubbles into the individual bubble objects after shadow removed as well. Bubble parameters, that is, bubble areas and bubble perimeters are obtained details, and results showed that the maximum errors between calculated results and measurement without considering shadow effects are approximately 25%. Thus, it indicated that the effects of shadow are seriously, and it must be optimized for better prediction regarding the ability and the capacity of reactor design.