Abstract Dynamic image processing techniques based on gray level were developed in this study to investigate the variation of two-phase flow patterns in the riser and its interaction with other thermal hydraulic behavior in a natural circulation loop. Four different modes of flows, namely, steady flow with flow rates less than 150 kg/h, periodic flow with a medium magnitude (peak value from 150 to 300 kg/h), periodic flow with a large magnitude (peak value greater than 300 kg/h), and chaotic flow were identified. Single-phase flow and bubbly flow with a very small void fraction (<0.2%) are the two predominant flow patterns in the lower riser, each of which takes about 50% of time, for both steady flow and periodic flow with a medium magnitude. Flow pattern variation for periodic flow with a large magnitude, which mainly takes place during double channel operation, is more complicated. For this mode of flow about 73–83% of time in the lower riser is of single-phase and bubbly flow with a very small void fraction, 4–8% of bubbly flow with a medium to high void fraction, 10–14% of slug flow and 3–8% of churn flow. The later two flow patterns appear periodically with the loop flow rate oscillation. The time percentage for each flow pattern in the lower rise for chaotic flow is quite sensitive to change of experimental conditions. The time percentage for single-phase flow and bubbly flow with a very small void fraction could be as high as 96%, but it could also be as low as 57%; for bubbly flow with a medium to high void fraction, slug flow and churn flow the time percentage ranges from 3 to 15%, 2 to 30% and 0 to 14%, respectively. Such simultaneous measurements of two-phase flow patterns and other thermal hydraulic properties enabled detail investigation into flow mechanism. Moreover, the single factor effect based on the Taguchi experiment matrix was also examined.