An Underwater Image Bubble Noise Removal Method Based on Optical Flow

It is very important for underwater robots to accurately detect and locate target objects. However, under many circumstances, it is difficult to clearly observe the target object due to the existence of bubble noise. In this paper, we proposed a method to solve this problem. First, we used the LK optical flow algorithm to calculate the motion vector of the image background and compensate for the background motion. Then, the optical flow field of the bubbles was calculated by the HS optical flow algorithm, and the area where the bubble existed was obtained by binarizing the image. Finally, we used the adjacent frame image to repair the bubble area. We carried out a bubble noise removal experiment. The results show that this method can effectively remove the bubble noise in the image.