Enhance Or Suppress Features enhances or suppresses certain image features (such as speckles, ring shapes, and neurites), which can improve subsequent identification of objects.
This module enhances or suppresses the intensity of certain pixels relative to the rest of the image, by applying image processing filters to the image. It produces a grayscale image in which objects can be identified using an Identify module.
Select the input image
Select the image with features to be enhanced or suppressed.
Name the output image
Enter a name for the feature-enhanced or suppressed image.
Select the operation
Select whether you want to enhance or suppress the features you designated.
- Enhance: Produce an image whose intensity is largely
composed of the features of interest.
- Suppress: Produce an image with the features largely
(Used only if circles, speckles or neurites are selected, or if suppressing features)
Enter the diameter of the largest speckle, the width of the circle
or the width of the neurites to be enhanced or suppressed, which
will be used to calculate an adequate filter size. To view pixel intensities in an open image, use the
pixel intensity tool which is available in any open display window. When you move
your mouse over the image, the pixel intensities will appear in the bottom bar of the display window.
(Used only if Enhance is selected)
This module can enhance three kinds of image intensity features:
- Speckles: A speckle is an area of enhanced intensity
relative to its immediate neighborhood. The module enhances
speckles using a white tophat filter, which is the image minus the
morphological grayscale opening of the image. The opening operation
first suppresses the speckles by applying a grayscale erosion to reduce everything
within a given radius to the lowest value within that radius, then uses
a grayscale dilation to restore objects larger than the radius to an
approximation of their former shape. The white tophat filter enhances
speckles by subtracting the effects of opening from the original image.
- Neurites: Neurites are taken to be long, thin features
of enhanced intensity. Choose this option to enhance the intensity
of the neurites using the Line structures or Tubeness methods
- Dark holes: The module uses morphological reconstruction
(the rolling-ball algorithm) to identify dark holes within brighter
areas, or brighter ring shapes. The image is inverted so that the dark holes turn into
bright peaks. The image is successively eroded and the eroded image
is reconstructed at each step, resulting in an image which is
missing the peaks. Finally, the reconstructed image is subtracted
from the previous reconstructed image. This leaves circular bright
spots with a radius equal to the number of iterations performed.
- Circles: The module calculates the circular Hough transform of
the image at the diameter given by the feature size. The Hough transform
will have the highest intensity at points that are centered within a ring
of high intensity pixels where the ring diameter is the feature size. You
may want to use the EnhanceEdges module to find the edges of your
circular object and then process the output by enhancing circles. You can
use IdentifyPrimaryObjects to find the circle centers and then use
these centers as seeds in IdentifySecondaryObjects to find whole,
circular objects using a watershed.
- Texture: EnanceOrSuppressFeatures produces an image
whose intensity is the variance among nearby pixels. This method weights
pixel contributions by distance using a Gaussian to calculate the weighting.
You can use this method to separate foreground from background if the foreground
is textured and the background is not.
- DIC: This method recovers the optical density of a DIC image by
integrating in a direction perpendicular to the shear direction of the image.
In addition, this module enables you to suppress certain features (such as speckles)
by specifying the feature size.
Range of hole sizes
(Used only if Dark holes is selected)
The range of hole sizes to be enhanced. The algorithm will
identify only holes whose diameters fall between these two
(Used only for the Texture, DIC or Neurites methods)
- Texture: This is the scale of the texture features, roughly
in pixels. The algorithm uses the smoothing value entered as
the sigma of the Gaussian used to weight nearby pixels by distance
in the variance calculation.
- DIC: Specifies the amount of smoothing of the image in the direction parallel to the
shear axis of the image. The line integration method will leave
streaks in the image without smoothing as it encounters noisy
pixels during the course of the integration. The smoothing takes
contributions from nearby pixels which decreases the noise but
smooths the resulting image.
- DIC: Increase the smoothing to
eliminate streakiness and decrease the smoothing to sharpen
- Neurites: The Tubeness option uses this scale
as the sigma of the Gaussian used to smooth the image prior to
Smoothing can be turned off by entering a value of zero, but this
is not recommended.
(Used only for the DIC method)
The shear angle is the direction of constant value for the
shadows and highlights in a DIC image. The gradients in a DIC
image run in the direction perpendicular to the shear angle.
For example, if the shadows run diagonally from lower left
to upper right and the highlights appear above the shadows,
the shear angle is 45°. If the shadows appear on top,
the shear angle is 180° + 45° = 225°.
(Used only for the DIC method)
The decay setting applies an exponential decay during the process
of integration by multiplying the accumulated sum by the decay
at each step. This lets the integration recover from accumulated
error during the course of the integration, but it also results
in diminished intensities in the middle of large objects.
Set the decay to a large value, on the order of 1 - 1/diameter
of your objects if the intensities decrease toward the middle.
Set the decay to a small value if there appears to be a bias
in the integration direction.
(Used only for the Neurites method)
Two methods can be used to enhance neurites:
- Tubeness: This method is an adaptation of
the method used by the
ImageJ Tubeness plugin. The image
is smoothed with a Gaussian. The Hessian is then computed at every
point to measure the intensity gradient and the eigenvalues of the
Hessian are computed to determine the magnitude of the intensity.
The absolute maximum of the two eigenvalues gives a measure of
the ratio of the intensity of the gradient in the direction of
its most rapid descent versus in the orthogonal direction. The
output image is the absolute magnitude of the highest eigenvalue
if that eigenvalue is negative (white neurite on dark background),
- Line structures: The module takes the difference of the
white and black tophat filters (a white tophat filtering is the image minus
the morphological grayscale opening of the image; a black tophat filtering is the
morphological grayscale closing of the image minus the image).
The effect is to enhance lines whose width is the "feature size".