Calvin asks about sub-pixel measuring.
Yes its possible. The key to doing this reliably is to understand the shape of the transition between the edge of the object and the background. A useful first approximation is to use a gaussian to model the shape of the transition. A paper by Kurt Benke some years ago explores this. (I am currently moving my office so I can't get an anything, so chasing the reference is left as an exercise). Essentially, the process will involve taking an image under controlled conditions of a well known object (standard). Producing an enlargement using either interpolation or not, try both. Plot the transition from black to white across the edges and choose the point which gives the correct answer. You the know where on the transition, for your conditions of lighting, contrast, camera etc, where the edge falls. The keeping everything the same, repeat the process with your unknowns. And it might sound easy but it is very difficult to keep everything the same. Best wishes Noel Goldsmith Noel Goldsmith Aircraft Forensic Engineering Air Vehicles Division DSTO 506 Lorimer Street Port Melbourne Vic 3207 AUSTRALIA Phone (613) 96267538 FAX (613) 96267089 Email [hidden email] > > > Date: Tue, 28 Mar 2006 07:20:04 -0800 > From: calope14 <[hidden email]> > Subject: Sub-Pixel Edge detection plugin? > > Hello to all, > I am doing an intensive project over the next couple weeks that is concerned > with computer vision dimensional guaging. I was planning to use ImageJ for > the image editing software as there are possibilities of needing to write a > plugin to gather and store data about a picture. > > But what I really need is a plugin that can find edges to a sub-pixel > standard. All I will be doing is finding the diameter of a circular object > to a sub-pixel standard. Does any one knnow of a plugin that does this. I > would gladly make one but this is an area that I know little about and dont > have time to do a lot of background learning. So if anyone could even let me > know the steps in the procedure of sub-pixel edge detection and the names of > some algorithms or processes used for them it would be a tremendous help, I > have searched the internet thouroughly and there are no real clear > insrtuctions for what I'm doing that doesnt cost money. I am sorry to bother > you all but time is really an issue so if anyone could give me a shove in > the right direction I would be most grateful. > Thank you for you time, and I wish you all luck on your current work > projects. > > Calvin Slope > -- > View this message in context: > http://www.nabble.com/Sub-Pixel-Edge-detection-plugin--t1356294.html#a3631519 > Sent from the ImageJ forum at Nabble.com. > |
Goldsmith, Noel writes:
>Calvin asks about sub-pixel measuring. > >Yes its possible. > ... >The keeping everything the same, repeat the process with your unknowns. >And it might sound easy but it is very difficult to keep everything the >same. Indeed. Also, consider this: before you spend a great deal of trouble trying to get sub-pixel measurements, be sure that you know the difference between precision and accuracy. In this group, it's tempting to bring up the concept of "empty magnification'. It's possible to "find" edge locations to whatever precision you like. The question is: are you gaining any ACCURACY? -- Kenneth Sloan [hidden email] Computer and Information Sciences (205) 934-2213 University of Alabama at Birmingham FAX (205) 934-5473 Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/ |
On Friday 31 March 2006 08:43, Kenneth Sloan wrote:
> Indeed. Also, consider this: before you spend a great deal of trouble > trying to get sub-pixel measurements, be sure that you know the > difference between precision and accuracy. > > In this group, it's tempting to bring up the concept of "empty > magnification'. > > It's possible to "find" edge locations to whatever precision you like. > The question is: are you gaining any ACCURACY? I think that one of the previous responses noted that you need an edge model. That is, you have to know in advance what the edge should look like. Once you have such a constraint, you can improve upon the classical resolution limit. Nothing magic here, it is just a way to optimally use prior knowledge. --aryeh -- Aryeh Weiss Dpmt of Electrical and Computer Engineering Bar Ilan University Ramat Gan 52900 Israel 972-3-5317722 ; 972-3-5340697 (FAX) |
Free forum by Nabble | Edit this page |