Interactive image segmentation based on geodesic active regions

Image segmentation is one of the basic and key technologies in image understanding and interpretation. In order to overcome the effect caused by noise, blur and background objects in image segmentation, in this paper an interactive image segmentation method, which combined prior knowledge and geodesic active regions, was proposed. Firstly, the area of target of interest was regarded as one class, and the remaining objects were regarded as background. Then the object and background were artificially labeled respectively, and through the selected features a probabilistic model was established. So the probabilities of each pixel belonged to the target area and the background area could be calculated respectively, and then the boundary and regional energy function were combined into the model to make the initial curve finally evolved to the target object boundary. This method had good performance in terms of speed and accuracy and had a strong ability to resist noise and blur.