Comments on: Have you heard of the “feature transform”? http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/ Steve Eddins manages the Image & Geospatial development team at <a href="http://www.mathworks.com/">The MathWorks</a> and coauthored <a href="http://www.mathworks.com/support/books/book5291.html?category=-1&language=-1">Digital Image Processing Using MATLAB</a>. He writes here about image processing concepts, algorithm implementations, and MATLAB.<br><br><img> Mon, 23 Nov 2009 00:50:01 +0000 http://wordpress.org/?v=2.3.1 By: Steve http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/#comment-21487 Steve Wed, 18 Feb 2009 02:16:44 +0000 http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/#comment-21487 Mark—Visualizing the feature transform using something like a quiver plot is a good idea. I'll put it on my list of blog ideas. As for filling holes smoothly, you should check out <tt>roifill</tt>, which does a kind of "soap-film" smooth interpolation from boundary pixels of arbitrary regions. Mark—Visualizing the feature transform using something like a quiver plot is a good idea. I’ll put it on my list of blog ideas. As for filling holes smoothly, you should check out roifill, which does a kind of “soap-film” smooth interpolation from boundary pixels of arbitrary regions.

]]> By: Mark Hayworth http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/#comment-21474 Mark Hayworth Sat, 14 Feb 2009 19:20:36 +0000 http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/#comment-21474 Steve, it's hard to visualize the L matrix but even more so because it uses linear indexing. But basically it's like a little vector pointing to the closest foreground pixel. It might be illustrative if we could visualize it like a vector field, like they use to show optical flow. Given that, potential applications might be easier to envision. For example, maybe it could be used to fill a hole in an object - with edge gray levels instead of uniform gray levels as is most common. The filled hole may look more smooth and natural than just a solid uniform patch. (Though this is probably not as smooth as a weighted average of all edge pixels.) Steve, it’s hard to visualize the L matrix but even more so because it uses linear indexing. But basically it’s like a little vector pointing to the closest foreground pixel. It might be illustrative if we could visualize it like a vector field, like they use to show optical flow. Given that, potential applications might be easier to envision. For example, maybe it could be used to fill a hole in an object - with edge gray levels instead of uniform gray levels as is most common. The filled hole may look more smooth and natural than just a solid uniform patch. (Though this is probably not as smooth as a weighted average of all edge pixels.)

]]> By: Oliver Woodford http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/#comment-21432 Oliver Woodford Thu, 29 Jan 2009 16:08:37 +0000 http://blogs.mathworks.com/steve/2009/01/28/have-you-heard-of-the-feature-transform/#comment-21432 Uncannily, I've just coded something that does roughly this for my print2im function. If only I'd known about bwdist! My problem is slightly different in that I don't care about background pixels that are too far away (> 4 pixels in x or y) from any foreground pixel. This allowed me to use convolution and powers of 2 to make it fast (without the IPT). Uncannily, I’ve just coded something that does roughly this for my print2im function. If only I’d known about bwdist!

My problem is slightly different in that I don’t care about background pixels that are too far away (> 4 pixels in x or y) from any foreground pixel. This allowed me to use convolution and powers of 2 to make it fast (without the IPT).

]]>