Abstract: Aiming at the bubble dynamic behaviors generated during the boiling flow in the rectangular narrow channel, the high-speed camera was used to capture the bubble behaviors, and then the image processing technology was used to bubble recognition, and finally the characteristic parameters are extracted for analysis. Through the improved image processing method, the problem that some unclosed bubble edges exist in the binarized image because of the light spots was solved, and a good gas-liquid segmentation effect was achieved. Then, the bubble projection area ratio of processed image was extracted by automatic processing, and the bubble projection area ratio by the manual recognition was used for verification, and the difference between the two is within 10%. Finally, the bubble volume calculation model was established according to the contact angle, and the change trend of void fraction during the bubble growth was obtained.