noise_distribution_mean ( ConstRegion, Image : : FilterSize : Distribution )

Determine the noise distribution of an image.

noise_distribution_mean calculates the noise distribution in a region of the image Image. The parameter ConstRegion determines a region of the image with approximately constant gray values. Ideally, the changes in gray values should only be caused by noise in this region. From this region the noise distribution is determined by using the mean__ operator to smooth the image, and to use the gray value differences in this area as an estimate for the noise distribution, which is returned in Distribution.


Attention

It is important to ensure that the region ConstRegion is not too close to a large gradient in the image, because the gradient values are then used for calculating the mean. This means the the distance of ConstRegion must be at least as large as the filter size FilterSize used for calculating the mean.


Parameters

ConstRegion (input_object)
region(-array) -> object
Region from which the noise distribution is to be estimated.

Image (input_object)
image -> object : byte
Corresponding image.

FilterSize (input_control)
integer -> integer
Size of the mean filter.
Default value: 21
Suggested values: 5, 11, 15, 21, 31, 51, 101
Range of values: 3 <= FilterSize <= 501 (lin)
Minimum increment: 2
Recommended increment: 2

Distribution (output_control)
distribution.values-array -> real
Noise distribution of all input regions.


Possible Predecessors

draw_region, circle, ellipse, rectangle1, rectangle2, threshold__, erosion_circle, gauss__, smooth__, sub__


Possible Successors

add_noise1, disp_distribution


See also

mean__, gauss_distribution



Copyright © 1996-1997 MVTec Software GmbH