Grid detection method
AssigneeAgfa Healthcare NV
InventorGert Behiels
Life of this patent
Estimating the term…
Abstract
Method of detecting an image of a grid in a digital image by extracting a number of features from the digital image and classifying extracted features in a classification algorithm, preferably implemented with support vector machines, which is trained with the same training features extracted from a number of reference images.
Claims22
A method of detecting a presence of an image of a linear grid parallel to one of the edges of an image represented by a digital image representation, comprising:
- computing predefined measures for different image locations in said image in a computer,
- determining an assumed direction of said grid by comparing said predefined measures for a first direction and a second direction,
- computing a confidence feature by steps including:
- deducing a reference value from the predefined measures in a direction perpendicular to said assumed direction,
- calculating said confidence feature as a ratio of a number of the predefined measures in the assumed grid direction which are larger than said reference value to the total number of predefined measures in the assumed direction,
- detecting a presence of said grid from the predefined measures and said confidence feature.
A method of detecting a presence of an image of a linear grid parallel to one of the edges of an image represented by a digital image representation comprising the steps of
- training a classification algorithm with classification parameters comprising predefined measures and at least one confidence feature extracted from a number of reference images in a computer,
- extracting said classification parameters from said digital image,
- classifying said classification parameters from said digital image in the trained classification algorithm,
- wherein said classification parameters comprise said measures and said at least one confidence features, said at least one confidence feature is computed by steps including:
- deducing a reference value from the predefined measures in a direction perpendicular to assumed direction of the grid,
- calculating said at least one confidence feature as a ratio of a number of the predefined measures in the assumed grid direction which are larger than said reference value to the total number of predefined measures in the assumed direction.
Dependencies are inferred from the claim text.