Interactive Graph Cuts for Optimal Boundary amp; Region Segmentation of Objects in N-D Images-英文文献.pdfVIP

Interactive Graph Cuts for Optimal Boundary Region Segmentation of Objects in N-D Images Yuri Y. Boykov Marie-Pierre Jolly Imaging and Visualization Department Siemens Corporate Research 755 College Road East, Princeton, NJ 08540, USA yuri @ Abstract mentation. A globally optimal segmentation can be very efficiently recomputed when the user adds or removes any In this paper we describe a new techniquefor general hard constraints (seeds). This allows the user to get any purpose interactive segmentation of N-dimensional images. desired segmentation results quickly via very intuitive in- The user marks certainpixels as “object”or “background” teractions. Our method applies to N-D images (volumes). to provide hard constraintsfor segmentation. Additional One of the main advantages of our interactive segmen- soji constraints incorporate both boundary and region in- tation method is that it provides a globally optimal solution formation. Graph cuts are used tofind the globally optimal for an N-dimensional segmentation when the cost function segmentation ofthe N-dimensional image. The obtained so- is clearly defined. Some earlier techniques (snakes [ 14,41, lution gives the best balance of boundary and regionprop- deformable templates [21], shortest path [ 151,ratio regions