Skip to Main Content
A new approach for image segmentation was proposed. In this method, a criterion named grade of geometric model consistency was introduced to determine what degree of objective region is in keeping with its prior geometric shape. Because of using different segmentation method or parameters to extract region will gain different result, and only one segmented outcome would optimally fit to its original model. So, in a sense, a region which has the highest grade of model consistency represents the best segmented outcome. However, key point of the whole approach was how to define the consistency and how to compute it. In this paper, the “consistency” definition and its related solution were brought forward and approach's two kind of concrete implementations were realized. Furthermore, in order to verify the validity of this segmentation method, an ultrasonic image of human heart was selected to extract the aorta region. Experiments illustrated that an objective region which has a certain prior geometric model could be segmented well by this means even if the original image's clarity is low or it affected by much noises.