Abstract:

A multi-region image segmentation method based on graph cuts is proposed. Original image data is transformed into high-dimension feature space via the implicit nonlinear mapping of data term by kernel function， so that the effect of segmentation is improved, multi-class partition of the image is achieved and the application of the piecewise constant model is extended. Due to the edge gradient of the image dramatically changes with discontinuity, the gradient constraint is introduced to smooth terms in order to reduce the over segmentation. Simultaneously, initial parameters are set by the unsupervised method without user interactions to meet the requirements of multi-region segmentation. Experiment results show that the proposed method is not restricted by the content of the image and has better segmentation results through both the subjective visual judgement and the quantitative analysis.