Intelligent scissors for image composition
read more
Citations
"GrabCut": interactive foreground extraction using iterated graph cuts
Computer Vision: Algorithms and Applications
Modeling and rendering architecture from photographs: a hybrid geometry- and image-based approach
Graphcut textures: image and video synthesis using graph cuts
Lazy snapping
References
A note on two problems in connexion with graphs
Snakes : Active Contour Models
Dynamic Programming
Theory of Edge Detection
Related Papers (5)
Interactive graph cuts for optimal boundary & region segmentation of objects in N-D images
Frequently Asked Questions (12)
Q2. What is the effect of cooling on the live-wire boundary?
The longer a pixel is on a stable section of the live-wire boundary, the cooler it becomes until it eventually freezes and automatically produces a new seed point.
Q3. How can training be used to train objects?
training can be turned on and off interactively throughout the definition of an object boundary so that it can be used (if needed) in a section of the boundary with similar edge characteristics and then turned off before a drastic change occurs.
Q4. How many seed points are required to ensure a closed object boundary?
Since each pixel (or free point) defines only one optimal path to a seed point, a minimum of two seed points must be placed to ensure a closed object boundary.
Q5. What is the effect of the formulation of boundary finding as a 2-D graph search?
formulation of boundary finding as a 2-D graph search eliminates the directed sampling and searching restrictions of previous implementations, thereby allowing boundaries of arbitrary com-G The authorx 2 The authory 2+=f Gm a x G( ) G− m a x G( )1
Q6. What is the purpose of equivalencing of spatial frequencies?
Equivalencing of spatial frequencies is performed by matching the spectral content of the cut piece and the destination image in the vicinity where it is to be pasted.
Q7. What are the important extensions of the work?
There are many rich extensions of this work, including: (1) making use of the weighted zero-crossings in the Laplacian to perform subpixel edge filtering and anti-aliasing, (2) use of multiple layered (multiplane) masks, (3) making spatial frequency equivalencing locally adaptive, (4) varying the light source over the object using directional gradient shading (artificial or borrowed) to provide consistent lighting in the composition, and, most importantly (5) extension of the 2-D DP graph search and application of the live-wire snap and training tools to moving objects and moving, multiplane masks for composition of image sequences.
Q8. What is the difference between training and a live-wire?
Since training is based on learned edge characteristics from the most recent portion of an object’s boundary, training is most effective for those objects with edge properties that are relatively consistent along the object boundary (or, if changing, at least change smoothly enough for the training algorithm to adapt).
Q9. What is the way to create a composition mask?
This requires the composition artist to “slip” the cutout object behind some scene components while leaving it in front of other components.
Q10. What is the difference between the livewire tool and the previous optimal contours?
the live-wire tool is much more similar to previous stage-wise optimal boundary tracking approaches than it is to snakes, since Intelligent Scissors were developed as an interactive 2-D extension to previous optimal edge tracking methods rather than an improvement on active contours.
Q11. How can The authorextend the gradient direction term without significant loss of computational efficiency?
it is possible to extend the gradient direction term to include 3 pixels and 2 links without significant loss of computational efficiency.
Q12. How can the object edge be estimated to subpixel accuracy?
That is, the position of the object edge can be estimated to subpixel accuracy by using a (linearly) weighted combination of the laplacian pixel values on either side of the zero-crossings.