Re: 3-connectivity in binary image

On Wed, 25 Feb 2009 17:46:38 -0800, ImageAnalyst wrote:
....
howard:
Yes, it's called convolution. Just convolve (a standard window scanning
technique) your binary image with a 3x3 kernel that is a cross shape (5
pixels on, and the corners off). The output image will be a gray scale
image, not a binary image. Then look at your output image and threshold
for intensities of 3 or 4 to find pixels that are 3 or 4 connected to
their neighbors. You can find convolution source code easily, I'm sure.

I never heard of 3-connectivity. What use does it have? What can it do
that the standard 4- and 8-connected don't do? Regards,
ImageAnalyst

I suppose it detects vertical or horizontal edges that are at least 2
pixels wide, while excluding 2x2 corners.

I have to wonder how valuable an edge detector it would be except on
images with *extremely* regular edges.

Randy
.

Relevant Pages

  • Re: 2D Sampling of images
    ... rather than computing the sum from each window position. ... the prior sum (at the prior window position). ... I have a question regarding how image pixels are sampled before I can ... carry out 2D convolution in the spatial domain. ...
    (sci.image.processing)
  • Re: None discrete convolution computation/arithmetic based on pixel blending and pixel registers.
    ... convolution based arithmetic feed into pixels, ... memorystacks constructed out of feedback looped cathodes to store ... There used to be talk of optical computing via programmable phase masks and diffraction, but I don't know how far that work has progressed. ...
    (sci.physics)
  • Re: Filtering kernel width
    ... , how wide should the window be, ... function leaves a realizable filter of finite size: ... becomes when it is convoluted with the function as more pixels ... Well I didn't see a problem with performing the convolution, ...
    (comp.graphics.algorithms)
  • Re: 3-connectivity in binary image
    ... source code for calculating 3-connectivity on a binary ... neighboring pixels. ... Yes, it's called convolution. ... The output image will ...
    (sci.image.processing)