Marine Images Segmentation Using Adaptive Fuzzy c-Means Algorithm Based on Spatial Neighborhood

In the marine environment, the underwater images have low SNR and the detail is fuzzy due to scattering and absorption of a variety of suspended matter in it and the water itself. If traditional methods are used to dispose underwater images directly, it is unlikely to obtain satisfactory results. Through the further research on traditional fuzzy c-means clustering algorithm, an adaptive FCM segmentation algorithm based on spatial information is proposed. The new method is based on the I₁I₂I₃ color space, and the characteristic distance in the target function is defined as the weighted sum between the distance of two pixels and spatial neighborhood distance, and an adaptive weight is given to the taking value method. By this new algorithm, the image spatial information is well taken into the FCM algorithm, and a new clustering objective function including neighborhood information is established, and thus effective marine image segmentation is achieved. Experimental results show that the new algorithm can achieve good quality of clustering and segmentation results, and has a strong ability to suppress noise.