Synchronous feature-tuning for underwater image segmentation

Image segmentation plays an important role for underwater object recognition. A new approach, called WA-SFT, which incorporates the watershed analysis and a synchronous feature-tuning (SFT) algorithm to perform fast underwater image segmentation, is presented. Currently, most watershed-based segmentation methods merge regions one by one to alleviate the over-segmentation problem. However, sequential merging would inevitably incur lengthy computation time. SFT simultaneously tunes features of regions by referring to adjacent regions. Due to the use of synchronous strategy, SFT achieves fast merging and provides great potentiality for a fully parallel hardware implementation. The iterative operation of WA-SFT converges when the numbers of merged regions in two successive iterations are identical. Empirical results show that WA-SFT outperforms other methods in terms of computation efficiency and segmentation accuracy.