System and method for shutter detection
View Patent ↗A method for removing shutter areas in an image, in particular an x-ray image is provided. Edges are examined in a multi-resolution image pyramid and evaluated to determine potential shutter blade candidates defining the shutter areas. Heuristic rules and/or an automatic classifier such as a Neuronal Network, are applied to distinguish true shutter blades from false positives. The rule set and the classifier are based on a set of features extracted from the potential shutter blade candidates as well as predetermined knowledge of the expected placement of the shutter, human anatomy. Up to four shutter blades are expected to be detected and based on these blades, the bright areas in the image that occur due to the shutters are removed.
1. A method for detecting and removing unwanted information in a medical image comprising:
processing said image to obtain edge information, wherein said processing comprises down sampling said image to generate a plurality of image levels having successively decreasing image resolutions;
generating a list of at least one candidate edge in said image from said edge information according to predetermined criteria, wherein each of said at least one candidate edge corresponds to a single blade of a shutter used to generate said image, the shutter including a plurality of blades, and wherein said list is generated by selecting a first candidate list at a lowest of said plurality of image levels, and promoting said first candidate list through subsequent image levels until a highest of said image levels and using candidate edges promoted to said highest of said image levels as said list of at least one candidate edge;
evaluating each of said at least one candidate edge according to a predetermined rule set to select one or more of said at least one candidate edge that is considered to define the extent of said unwanted information within said image, said evaluating comprising comparing said at least one candidate edge to other information in said image; and
removing said unwanted information based on the location of said one or more of said at least one candidate edge in said image.
2. The method according to claim 1 wherein said at least one candidate edge is chosen by applying a score function to edges detected in said image.
3. The method according to claim 1 wherein said evaluating each of said at least one candidate edge comprises evaluating properties of one or more segments for each said at least one candidate edge.
4. The method according to claim 1 wherein said evaluating each of said at least one candidate edge comprises evaluating properties of candidate edges according to the orientation of said candidate edges in said image, said properties being those associated with an image side of said candidate edges closer to a center of said image and a shutter side of said candidate edges further away from said center of said image.
5. The method according to claim 1 wherein said predetermined rule set is prepared using one or more classifiers applied during a training phase.
6. The method according to claim 1 wherein said evaluating each of said at least one candidate edge comprises evaluating pairs of candidate edges having similar orientations.
7. The method according to claim 1 wherein said evaluating each of said at least one candidate edge comprises considering predetermined properties of anatomy to discard false positives.
8. A computer readable medium carrying computer executable instructions for causing an image processing device to execute the method according to claim 1 .
9. A system for detecting and removing unwanted information in a medical image comprising:
an interface comprising a display for viewing said image and enabling a user to interact with said system; and
an image processing program capable of obtaining and processing said image from a medical imaging device and displaying said image on said display, said image processing program being configured for
processing said image to obtain edge information, wherein said processing comprises down sampling said image to generate a plurality of image levels having successively decreasing image resolutions;
generating a list of at least one candidate edge in said image from said edge information according to predetermined criteria, wherein each of said at least one candidate edge corresponds to a single blade of a shutter used to generate said image, the shutter including a plurality of blades and wherein said list is generated by selecting a first candidate list at a lowest of said plurality of image levels, and promoting said first candidate list through subsequent image levels until a highest of said image levels and using candidate edges promoted to said highest of said image levels as said list of at least one candidate edge;
evaluating each of said at least one candidate edge according to a predetermined rule set to select one or more of said at least one candidate edge that is considered to define the extent of said unwanted information within said image, said evaluating comprising comparing said at least one candidate edge to other information in said image; and
removing said unwanted information based on the location of said one or more of said at least one candidate edge in said image.
10. The system according to claim 9 wherein said at least one candidate edge is chosen by applying a score function to edges detected in said image.
11. The system according to claim 9 wherein said evaluating each of said at least one candidate edge comprises evaluating properties of one or more segments for each said at least one candidate edge.
12. The system according to claim 9 wherein said evaluating each of said at least one candidate edge comprises evaluating properties of candidate edges according to the orientation of said candidate edges in said image, said properties being those associated with an image side of said candidate edges closer to a center of said image and a shutter side of said candidate edges further away from said center of said image.
13. The system according to claim 9 wherein said predetermined rule set is prepared using one or more classifiers applied during a training phase.
14. The system according to claim 9 wherein said evaluating each of said at least one candidate edge comprises evaluating pairs of candidate edges having similar orientations.
15. The method according to claim 1 wherein said evaluating each of said at least one candidate edge comprises considering predetermined properties of anatomy to discard false positives.