median_weighted ( Image : ImageWMedian : MaskType, MaskSize : )

Weighted median filtering with different rank masks.

The operator median_weighted calculates the median of the gray values within a local environment. In contrast to median, which uses all gray values within the environment exactly once, the operator median_weighted weights all gray values several times depending on their position. A gray value is received into the field to be sorted several times according to its weighting. The following masks are available:

'gauss' at MaskSize = 3

1 2 1 
2 4 2
1 2 1

'inner' at MaskSize = 3

1 1 1
1 3 1
1 1 1
The operator median_weighted means that, contrary to median, gray value corners remain.


Parameters

Image (input_object)
image(-array) -> object : byte / int2
Image to be filtered.

ImageWMedian (output_object)
image(-array) -> object : byte / int2
Median filtered image.

MaskType (input_control)
string -> string
Type of median mask.
Default value: 'inner'
List of values: 'inner', 'gauss'

MaskSize (input_control)
integer -> integer
mask size.
Default value: 3
List of values: 3


Example
read_image(:Image:'fabrik':) >
median_weighted(Image:MedianWeighted:'gauss',3:) > 
disp_image(MedianWeighted:::).

Complexity

For each pixel: O(F * log(F)) with F = area of MaskType.


Possible Predecessors

read_image


Possible Successors

threshold__, dyn_threshold__, regiongrowing__


Alternatives

median, trimmed_mean, sigma__


References

R. Haralick, L. Shapiro; "Computer and Robot Vision"; Addison-Wesley, 1992, Seite 319



Copyright © 1996-1997 MVTec Software GmbH