A clustering-plane based energy optimization method for image segmentation

A novel regional energy minimization model is proposed in this paper, which aims to find the optimal clustering planes for respective objects in the image domain. By using the distances from the pixels to the center planes and spatial location information, the model assigns the pixels to the appropriate categories. A soft membership function is introduced to estimate the score which describes the possibility that the pixel falls into the category. Further, the spatial information is employed to amend the membership function so as to enhance the noise robustness of the model. The parameters of the center planes are updated through the energy minimization, and constrained by the prior values as well. The proposed model has been conducted on the synthetic images and real images, quantitatively and qualitatively, to demonstrate its performance.