Long-distance sea wave extraction method based on improved Otsu algorithm

Sea wave extraction is the foundation of sea wave measurement, which plays a significant role in binocular stereo vision-based tsunami early-warning systems. However, long-distance sea surface images possess two typical features: ununiformed illuminance and a low signal-to-noise ratio that makes it difficult to extract sea waves out from sea surface images and that leads to incorrect extractions. In this paper, we improved the Otsu algorithm to realize a dynamic multi-threshold algorithm for sea wave extraction. First, the original image was divided into small blocks and the dynamic multi-thresholds were determined according to local sea surface illuminance. Then an evaluation function was defined to add constraint conditions to adjacent blocks' thresholds. Finally, sea waves were extracted out using the blocks' thresholds in all the blocks. As seen in our experimental results, the correct extraction rate was 80.9%, most sea waves are extracted regardless of their intensity.