points_lepetit🔗
Short description🔗
points_lepetit — Detect points of interest using the Lepetit operator.
Signature🔗
points_lepetit( image Image, integer Radius, integer CheckNeighbor, integer MinCheckNeighborDiff, integer MinScore, string Subpix, out point.y Row, out point.x Column )
Description🔗
points_lepetit extracts points of interest like corners or
blob-like structures from Image. The Image is first
smoothed with a median of size 3x3. Then, all the gray values on a circle
with radius Radius around an interest point candidate
(m) are examined. The absolute differences of two diagonally
opposed gray values (m1,m2) on the circle to the central pixel
m is computed. At least one of these differences has to be
larger than MinCheckNeighborDiff. All diagonally opposed pixels
on the circle must fulfill that condition. To suppress detection of
points at edges that have a small curvature (aliasing), it is possible
to compute CheckNeighbor further differences of circle point
neighbors of m1 and m2 to the center, that as well
fulfill the above criteria. By computing all gray value differences of
the circle points to the center, a mean gray value difference is
determined. That value has to be larger than MinScore and
allows to restrict the results to points with high contrast. By computing
the score of all eight neighbors of m, it is possible to fit a quadratic
equation to that. The maxima of that equation determines a subpixel
accurate interest point position. By setting the parameter Subpix
to ‘interpolation’ (default) or ‘none’, it is possible to turn that refinement
step on or off. The resulting points are returned in Row and
Column. The operator points_lepetit can especially be
used for very fast interest point extraction. The results are however
less robust than points extracted by points_harris for example.
Attention🔗
Note that filter operators may return unexpected results if an image with a reduced domain is used as input. Please refer to the chapter Filters.
Execution information🔗
Execution information
-
Multithreading type: reentrant (runs in parallel with non-exclusive operators).
-
Multithreading scope: global (may be called from any thread).
-
Automatically parallelized on domain level.
Parameters🔗
Image (input_object) singlechannelimage → object (byte / uint2)
Input image.
Radius (input_control) integer → (integer)
Radius of the circle.
Default: 3
Suggested values: 3, 5, 6, 7, 8, 9, 10, 15
CheckNeighbor (input_control) integer → (integer)
Number of checked neighbors on the circle.
Default: 1
Suggested values: 1, 2, 3, 5
MinCheckNeighborDiff (input_control) integer → (integer)
Threshold of gray value difference to each circle point.
Default: 15
Suggested values: 10, 15, 20, 25, 30, 35, 40, 45, 60, 80
MinScore (input_control) integer → (integer)
Threshold of gray value difference to all circle points.
Default: 30
Suggested values: 5, 10, 15, 20, 25, 30
Subpix (input_control) string → (string)
Subpixel accuracy of point coordinates.
Default: 'interpolation'
List of values: 'interpolation', 'none'
Row (output_control) point.y-array → (integer / real)
Row-coordinates of the detected points.
Column (output_control) point.x-array → (integer / real)
Column-coordinates of the detected points.
Combinations with other operators🔗
Combinations
Possible predecessors
Alternatives
points_foerstner, points_harris, points_harris_binomial, points_sojka
Module🔗
Foundation