Abstract: Abstract: In order to overcome the video data redundancy and image haze in the video capture process during the underwater detection, this paper designs a keyframe extraction system taking into account the video frame content and extract amount. The original detective image is enhanced by using the Image Dehazing Algorithm based on Dark Channel Prior. The improved key frame extraction algorithm based on sampling interval and image similarity is used for keyframe extraction after the original image enhancement. The image stitching method is used to test the reliability of extracted key frames. Experimental results show that the amount of key frames is significantly reduced comparing to the original data, and effectively reflects the main information for the entire detective process.

Key words: keyframe, defogging, image enhancement, sampling interval, image stitching