• Next message: Brian Kuyatt: "RE: ImagePro> 16 bit gray scale conversion"
• Previous message: michael shaffer: "RE: ImagePro> 16 bit gray scale conversion"
• Maybe in reply to: Jill Schmidt: "ImagePro> how do dilate/erode work?"
• Next in thread: Vern: "RE: ImagePro> count/measure"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

"Kevin Ryan" <kevin@mediacy.com> writes:

Dilate and erode are nonlinear morphological filters, which means that the
resulting pixel will be some nonlinear result of the local neighborhood, and
that the result is dependent on the shape of the neighborhood examined.

Your description of the filters is correct. Assuming a binary image: in a
3x3 dilate operation, if any pixel in the 3x3 region (plus or minus a pixel
in any direction) is white, the resulting central pixel is white. The effect
of this is to enlarge white objects by expanding their borders. In a 3x3
erosion, if any of the pixels in the 3x3 region is black, the output pixel
is black. The effect here is to shrink the white objects by 'eroding' their
edges. Only the center pixel changes color during this operation.

3x3 cross, 3x3 square, 11x11 octagon, etc. are all descriptions of the
region checked for black (or white) pixels. Note that the larger the region,
the more pixels will change value in the output image. The shape of the
neighborhood will affect the results of the morphological operation.

Grayscale dilation and erosion work in a slightly different manner:
grayscale dilation makes the central pixel equal to the brightest pixel in
the neighborhood, while erosion makes it equal to the darkest. Again,
dilation causes bright objects to grow, and erosion causes them to shrink.

In a related manner, morphological opening is an erosion followed by a
dilation, while closing is a dilation followed by an erosion. Opening tends
to break thin lines and remove small 'hair' around objects, while closing
can join nearby objects and fill small holes. Binary thinning
(skeletonization) is similar to an erosion, with additional conditions to
prevent removing pixels that affect the connectivity and general shape of
the original object.

Books such as John Russ's "Image Processing Handbook" or Ken Castlemans
"Digital Image Processing" cover these topics in more detail, and are
excellent resources.

Hope this is helpful!

-- Kevin Ryan
kevin@mediacy.com

-----Original Message-----
From: Jill Schmidt <jills@ocean.washington.edu>
To: imagepro-users@mediacy.com <imagepro-users@mediacy.com>
Date: Tuesday, May 25, 1999 2:24 PM
Subject: ImagePro> how do dilate/erode work?


>Jill Schmidt <jills@ocean.washington.edu> writes:
>
>
>Hi,
>Can someone please tell me how these filters actually work?  For
>simplicity, assume a binary image, white grains on black background.
>
>For dilate, does the center value of the kernel change to white if any
>part of the kernel is on a white pixel?  Or does the whole kernel turn
>white if any part of the kernel is on a white pixel?  Is erode the
>reverse, i.e. if any part of the kernel is on a black pixel it turns the
>center pixel black?
>
>Thanks!!  Jill
>

---------------------------------
To unsubscribe from this list, send an email to majordomo@mediacy.com with
'unsubscribe imagepro-users' as the body of the message.

• Next message: Brian Kuyatt: "RE: ImagePro> 16 bit gray scale conversion"
• Previous message: michael shaffer: "RE: ImagePro> 16 bit gray scale conversion"
• Maybe in reply to: Jill Schmidt: "ImagePro> how do dilate/erode work?"
• Next in thread: Vern: "RE: ImagePro> count/measure"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This archive was generated by hypermail 2b29 : 08/04/00 EDT