An improved K-means algorithm for underwater image background segmentation

Abstract Conventional algorithms fail to obtain satisfactory background segmentation results for underwater images. In this study, an improved K-means algorithm was developed image address the issue of improper K value determination and minimize impact initial centroid position grayscale during gray level quantization conventional algorithm. A total 100 images taken by robot were sampled test aforementioned in respect validity time cost. The optimized. compared other three existing algorithms, including algorithm, Otsu Canny operator edge extraction method. experimental showed that could effectively segment with a low color cast, contrast, blurred edges. Although its cost higher than it none less proved more efficient time-consuming manual proposed paper potentially be used environments segmentation.