LabVIEW
Fitting a surface (height=f(x,y) with a least square fit
Could you repeat the question, it is too ambiguous.
What do you want to "remove"? Are you talking about masking a circular area?
Is your 2D array a (2,x) array, wich each row describing the x,y coordinates of a point? In this case, you can only do a circle because there is no 3rd dimension.
Do you have a full 2D grid with values representing the z-direction (similar to an intensity graph)? In this case, you can probably fit to e.g. a half-sphere after you define the the points belonging to it. As CC mentioned, fitting on a sphere should do the trick.
Attach an example of your data and we can help further.
What do you want to "remove"? Are you talking about masking a circular area?
Is your 2D array a (2,x) array, wich each row describing the x,y coordinates of a point? In this case, you can only do a circle because there is no 3rd dimension.
Do you have a full 2D grid with values representing the z-direction (similar to an intensity graph)? In this case, you can probably fit to e.g. a half-sphere after you define the the points belonging to it. As CC mentioned, fitting on a sphere should do the trick.
Attach an example of your data and we can help further.
Hey, cool. I thought I recognized that 2D surface fit ;). (Maybe one of these days I should submit it to the example code library here on ni.com...)
My green Lev-Mar subVIs are a dramatic improvement in functionality over the stock Vis where the function needs to be called one point at a time. (These are actually highly simplified, my latest internal versions are much more universal and can do all kinds of tricks :))
Levenberg-Marquardt is definitely the way to go if you want to subtract an arbitrary 2D background surface from your data.
IF your data really is an exactly spherical surface as you (maybe?) first hinted, it would be much more efficient to use the "fitting on a sphere". It uses a very interesting algorithm that reduces the problem to a simple linear equation (see the online help). I made a quick demo last night while watching news. For simplicity, it assumes that the center of the sphere is at z=0, modify as needed. (See attached, LabVIEW 7.1)
My green Lev-Mar subVIs are a dramatic improvement in functionality over the stock Vis where the function needs to be called one point at a time. (These are actually highly simplified, my latest internal versions are much more universal and can do all kinds of tricks :))
Levenberg-Marquardt is definitely the way to go if you want to subtract an arbitrary 2D background surface from your data.
IF your data really is an exactly spherical surface as you (maybe?) first hinted, it would be much more efficient to use the "fitting on a sphere". It uses a very interesting algorithm that reduces the problem to a simple linear equation (see the online help). I made a quick demo last night while watching news. For simplicity, it assumes that the center of the sphere is at z=0, modify as needed. (See attached, LabVIEW 7.1)
Message Edited by altenbach on 03-24-2005 08:51 AM
