Bubbles extraction in the images of aerated water flows accurately is the precondition to analyze the characteristics of water flows by image detection. Because bubbles are colorless and transparent in the aerated water flows and the contrast of object and background is too low and at the same time there are un-even illuminations during experimentation. It is difficult to segment bubbles accurately. In order to solve the problems of miss-extraction or deformity of bubbles which caused by un-uniformity of bubbles resulted from the un-balance illumination of the image

a method based on two layer dynamic thresholding is proposed in this paper. At first

the local areas which containing bubbles are obtained dynamically by labeling in the first layer processing. Then

bubbles are extracted accurately in the second layer with fine segmentation based on the labeled bubbles' area in the first layer. Experimental results show that the proposed method is very efficient in solving the bad illumination and can highly improve the accuracy of bubble extraction.