An improved KFCM algorithm for unsupervised image segmentation based on neighborhood constraints

Fuzzy clustering has been widely applied in the field of image segmentation, but it is sensitive for the noise in image. To solve this problem an unsupervised image segmentation improvement scheme combining neighborhood spatial constraints is proposed in this paper. The objective function of traditional fuzzy C-means clustering in the scheme are modified, and the kernel function and the neighbor spatial information are combined to improve the mode of action of the neighboring pixels on the center pixel, so that the center pixel of window can be adaptively updated through neighborhood pixel to achieve the purpose of filtering noise. The algorithm is applied and tested in synthetic and real images with salt and pepper noise and Gaussian noise, and the experimental results shown that compared with the other five traditional fuzzy C-means clustering and their improved schemes, the proposed algorithm is robust to noise, and the segmentation accuracy is significantly improved. In addition, the fuzzy clustering performance of the algorithm is also improved in the validity of the fuzzy division tested by three indicators.