Comments on: Linear indexing http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/ 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> Sun, 08 Nov 2009 02:28:13 +0000 http://wordpress.org/?v=2.3.1 By: Steve http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20312 Steve Sat, 23 Feb 2008 19:50:12 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20312 JanKees—See <tt>normxcorr2</tt>. JanKees—See normxcorr2.

]]> By: JanKees http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20307 JanKees Sat, 23 Feb 2008 12:37:30 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20307 Hi, Steve: As you are dealing with indexing, maybe my question fits in this topic: is there a fast way to do template matching in Matlab? I am thinking in something related to xcorr/xcorr2 or something related to it (as in the case of the summation of the absolute/squared difference). Best regards from, JanKees Hi, Steve:

As you are dealing with indexing, maybe my question fits in this topic: is there a fast way to do template matching in Matlab? I am thinking in something related to xcorr/xcorr2 or something related to it (as in the case of the summation of the absolute/squared difference).

Best regards from,

JanKees

]]> By: Steve http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20210 Steve Sat, 16 Feb 2008 13:16:45 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20210 Kevin—The later chapters are about image analysis and object recognition. There's nothing in the book about tracking. Kevin—The later chapters are about image analysis and object recognition. There’s nothing in the book about tracking.

]]> By: Kevin Lan http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20206 Kevin Lan Sat, 16 Feb 2008 07:28:10 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20206 They are not quite parallel but I will try that, thank you. And I just bought your book "Digital Image Processing using Matlab" to try to solve one of my recent projects. Which part of the book should I read to solve pattern tracking/detection/recognition problems? I basically have a light microscope image that has some fingerprint-like patterns that I would like to detect and mark. Or what methods would you recommend? Thank you! They are not quite parallel but I will try that, thank you. And I just bought your book “Digital Image Processing using Matlab” to try to solve one of my recent projects. Which part of the book should I read to solve pattern tracking/detection/recognition problems? I basically have a light microscope image that has some fingerprint-like patterns that I would like to detect and mark. Or what methods would you recommend? Thank you!

]]> By: Steve http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20201 Steve Fri, 15 Feb 2008 20:27:41 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20201 From your description, especially the "lines very much blended to the background color" part, it sounds challenging. If the lines are parallel, however, you might try computing a projection along the direction parallel to the lines. The line locations would show up as peaks in the projection. From your description, especially the “lines very much blended to the background color” part, it sounds challenging. If the lines are parallel, however, you might try computing a projection along the direction parallel to the lines. The line locations would show up as peaks in the projection.

]]> By: Kevin Lan http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20182 Kevin Lan Thu, 14 Feb 2008 20:20:31 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20182 Hello Steve, I am trying to analyze a scanning electron microscope image which has many layers bounded very closely which makes the edges/boundaries of the layers very unclear (but can still tell by naked eyes). It also has some obvious cracks as well. I tried a couple of edge functions and they didn't give me satisfactory results, either too much distaction on surrounding or no edges except for cracks. If you have hard time imagining the image, it just looks a page of lined notepad with the lines very much blended to the background color and with a few obvious crack spreaded on it. Is there a way that I can detect the edges accurately and clearly? Thank you! Hello Steve,
I am trying to analyze a scanning electron microscope image which has many layers bounded very closely which makes the edges/boundaries of the layers very unclear (but can still tell by naked eyes). It also has some obvious cracks as well. I tried a couple of edge functions and they didn’t give me satisfactory results, either too much distaction on surrounding or no edges except for cracks. If you have hard time imagining the image, it just looks a page of lined notepad with the lines very much blended to the background color and with a few obvious crack spreaded on it. Is there a way that I can detect the edges accurately and clearly?
Thank you!

]]> By: Steve http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20180 Steve Thu, 14 Feb 2008 13:58:53 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20180 Tim—Thanks for the clarification. Almost all of my image processing work is with dense matrices. Tim—Thanks for the clarification. Almost all of my image processing work is with dense matrices.

]]> By: Tim Davis http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20178 Tim Davis Thu, 14 Feb 2008 13:49:02 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20178 Linear indexing is great for dense matrices, but use it with caution for sparse matrices. It's very slow (the problem is intrinsically difficult), and it breaks down if prod(size(A)) is greater than 2^31. It's quite easy to work with sparse matrices that large, but not with linear indexing. It would be really easy to implement a kind of indexing where you could refer to the kth nonzero element in the matrix (in column order). That is, the kth nonzero would be the kth entry in i, j, and x where [i j x] = find (A), not the A(row,col) index where k=M*(col-1)+row. That would be extremely fast, as compared with linear indexing based on row/column indices, but M doesn't support that kind of indexing. Linear indexing is great for dense matrices, but use it with caution for sparse matrices. It’s very slow (the problem is intrinsically difficult), and it breaks down if prod(size(A)) is greater than 2^31. It’s quite easy to work with sparse matrices that large, but not with linear indexing.

It would be really easy to implement a kind of indexing where you could refer to the kth nonzero element in the matrix (in column order). That is, the kth nonzero would be the kth entry in i, j, and x where [i j x] = find (A), not the A(row,col) index where k=M*(col-1)+row. That would be extremely fast, as compared with linear indexing based on row/column indices, but M doesn’t support that kind of indexing.

]]> By: Daisuke Aoyagi http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20124 Daisuke Aoyagi Tue, 12 Feb 2008 19:45:29 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20124 I get confused between col-order (matrix) and row-order (subplot) sometimes. I thought it'd be cleaner (less confusing) if they are consistent, plus it'd be nice to be able to use something built-in like sub2ind to specify col & row of subplots, instead of N*(row-1)+col (and mess up). Not a big deal, though. I get confused between col-order (matrix) and row-order (subplot) sometimes. I thought it’d be cleaner (less confusing) if they are consistent, plus it’d be nice to be able to use something built-in like sub2ind to specify col & row of subplots, instead of N*(row-1)+col (and mess up). Not a big deal, though.

]]> By: Steve http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20094 Steve Tue, 12 Feb 2008 02:57:41 +0000 http://blogs.mathworks.com/steve/2008/02/08/linear-indexing/#comment-20094 Daisuke—I have no idea. That behavior was established probably 20 years ago. I've been at MathWorks at long time but not that long. Why does it matter to you? Daisuke—I have no idea. That behavior was established probably 20 years ago. I’ve been at MathWorks at long time but not that long. Why does it matter to you?

]]>