Method for determining optimal chamfer mask coefficients for distance transform
AssigneeThales
InventorsElias Bitar, Nicolas Marty
Life of this patent
Estimating the term…
Loading patent…
AssigneeThales
InventorsElias Bitar, Nicolas Marty
Estimating the term…
The coefficients of a chamfer mask are, to within a multiplicative scale factor making it possible to give them an integer value, approximations of the Euclidian distances separating the pixels covered by the mask, from the pixel under analysis placed at the center of the mask. As there are at least two possible integer values for each coefficient, the over- and the under-approximation, one is rapidly faced with a considerable number of possible combinations. The method proposed allows progressive selection of the possible integer values, firstly at the level of each coefficient by virtue of an axis error rate criterion, then at the level of the coefficients considered by binomials by virtue of a sector error rate criterion, which considerably reduces the number of combinations to be analyzed to arrive at an optimal combination from the point of view of the error rate obtained in the distance estimations.
A method for determining optimal coefficients Ci,j, i being a row index varying from −P to +P and j a column index varying from −Q to +Q, P and Q being positive integers, of a distance transform chamfer mask providing estimations of distance between pixels of an image, said coefficients being approximations, to within a multiplicative scale factor n, n being a positive integer, of the distances of an image pixel subjected to the analysis of the chamfer mask with respect to the image pixels that are closest in the various so-called pixel directions of the mask, each coefficient being associated with the pixel of the mask whose distance it represents, and the determination consisting of a choice, for each coefficient, of a value selected from among a set of integer candidate values for the coefficient considered, said method comprising:
Dependencies are inferred from the claim text.