Process > Filters-2D Filters > Edge

 

The Image-Pro edge filters are available through the 2D Filters dialog's Edge tab.

To open the 2D Filters dialog, find the Filters group on the Adjust tab’s ribbon, click on the down arrow, and then click on 2D Filters.

The available edge filters are described below.

Sobel:   Select this filter if you want to enhance just the principal edges in an image.  The Sobel applies a mathematical formula to a 3 x 3 neighborhood to locate and highlight its edges.

Roberts:   Select this filter if you want to enhance fine edges in an image.  The Roberts filter is not a convolution filter.  It applies a mathematical formula upon a 4 x 4 neighborhood to produce its effect.  The upper left pixel in the neighborhood is the one replaced.

Laplacian: Select this filter if you want to enhance all the edges in an image.  See below for a description of the controls you can use to 'tune' this filter.

Variance:   Select this filter if you want to detect and emphasize edges and textures.  The Variance filter substitutes a pixel with the standard deviation for its neighborhood. If you select Variance, the Width and Height controls are displayed.  These determine the kernel size to be used.  A small kernel produces a more subtle filtering effect.  A large kernel produces a more extreme filtering effect.

Sobel Phase: Select this filter if you want to enhance edges in a manner that also indicates the direction of the intensity change.  The Phase filter is the complement of the Sobel filter.

Horizontal Edges: Use this filter to detect and emphasize horizontal edges.  See below for a description of the controls you can use to 'tune' this filter.

Vertical Edges: Use this filter to detect and emphasize vertical edges.  See below for a description of the controls you can use to 'tune' this filter.

Canny: Select this filter if you want to apply the Canny edge detection filter.  This filter can be used to extract gradient edges. The Canny edge detection filter algorithm includes the following steps:

  1. Noise reduction - A Gaussian filter of the specified Kernel size is applied to reduce image noise.
  2. Edge detection - Edges are detected using a Sobel filter. The orientation of edges is also detected by applying a SobelPhase filter.
  3. Non-maxima suppression - The neighbors of each pixel in the normal direction to the edge are analyzed and if one of the neighbors is bigger that the pixel value, the pixel is suppressed.
  4. (Optional) Hysteresis thresholding - If you select the Hysteresis thresholding check box, two thresholds are applied to the image. If a pixel value is less than or equal to the value you enter for Low, then the pixel is labeled as weak and excluded from the edge; if a pixel value is greater than the value you enter for High, then the pixel is labeled strong and included in the edge; the rest of the pixels are considered "candidates." At the Edge linking stage, all candidate pixels which are connected to strong pixels are included in the edge.  

This filter is applied to the luminance channel in the case of color images.
 

Gabor: The Gabor filter is a linear filter whose impulse response is defined by a harmonic function multiplied by a Gaussian function. The harmonic function is defined by orientation and scale (wavelength). The result is a new floating point image.

  • Period – Period of the frequency harmonic (in pixels) used by the Gabor filter
  • Angle - Defines the Phase offset of the filter in degrees.
  • Absolute Output - Defines whether the result image of Gabor filtration uses the Abs operation.

DIC restore: Select this filter to restore a Differential Interference Contrast (DIC) image, converting interference contrast to density.  If you select this option, the Angle control appears.  Use the Angle control to specify the angle of periodic features to be enhanced by the filter. Angle is measured from the horizontal direction clockwise in degrees

If you have selected Laplacian, Horizontal Edges or Vertical Edges

If you have selected the edge filters The , the options will relate to kernel size and filtering strength. The following options will be presented:

  • 3 x 3:   Select this button to use the 3 x 3 kernel size.  A small kernel produces a more subtle filtering effect.
  • 5 x 5:   Select this button to use the 5 x 5 kernel size.
  • 7 x 7:   Select this button to use the 7 x 7 kernel size.  A large kernel produces a more extreme filtering effect.
  • Passes:   Enter the number of times you want the filter applied to your image.  When a filter is applied multiple times, its effect is amplified by each pass.  An image that has been softened by one pass of the LoPass filter, will be softened further by a second pass.
  • Strength:   Enter a value from 1 - 10 that reflects how much of the filtering effect you want applied to the image.  A value of 10 specifies that you want the full strength (100%) of the filtered result applied to each pixel.  Values less than 10 cut the full weight of the filter - a value of 1 indicates that only 10% of the difference between the filtered pixel value and the original pixel value should be applied, a value of 2 indicates that 20% of the difference should be applied, and so forth.

If you have selected Sobel, Sobel_Phase or Roberts

If you have selected Sobel, Sobel Phase or Roberts, no options are available.

If you have selected Variance

If you have selected Variance, the Width and Height options are available.  These determine the kernel size to be used.  A small kernel produces a more subtle filtering effect.  A large kernel produces a more extreme filtering effect.