Simple image processing question?

Share icon

Wed, 06/22/2016 - 03:43 am

I want to open an image file (simple via Image Load) and use ImageInterploate to explore the effect of Voronoi tessellations. I need to make the image wave into a triplet wave, but I can't find out how to do this. I presume I then need to convert the tessellated wave back into an image. Is this a sensible thing to try and if so, how should I go about it? DN
I couldn't find a straightforward way to convert an image wave to a triplet wave, but it's not difficult to use some "for" loops for this purpose. Presumably there's some clever MatrixOP or ImageTransform operation that can do it more efficiently. Here's some code that converts the image to a triplet wave then back and plots it; somewhere along the way I lost a pixel, but it's close. Add your tessellation code in the middle.

function imagetoxyz(imgwave)
    //operates on a 2d source wave, for example grayscale
    wave imgwave

    //first convert the image to a triplet wave using "for" loops
    make/O/N=(numpnts(imgwave),3) xyzwave
    variable i=0,j=0,k=0
    for(i=0;i<dimsize(imgwave,0);i+=1)
        for(j=0;j<dimsize(imgwave,1);j+=1)
            xyzwave[k][0]=i
            xyzwave[k][1]=j
            xyzwave[k][2]=imgwave[i][j][k]
            k+=1
        endfor
    endfor
   
    ///
    //Do your voroni tesselation processing here on the triplet wave "xyzwave"
    ///
   
    //convert back to a standard image, then plot:
    Make/O/N=(xyzwave[DimSize(xyzwave,0)-1][0],xyzwave[DimSize(xyzwave,0)-1][1]) newimgwave
    duplicate/O newimgwave countMat
    ImageFromXYZ xyzwave, newimgwave, countMat
    NewImage newimgwave
end

June 22, 2016 at 09:19 am - Permalink

ajleenheer wrote:
I couldn't find a straightforward way to convert an image wave to a triplet wave, but it's not difficult to use some "for" loops for this purpose. Presumably there's some clever MatrixOP or ImageTransform operation that can do it more efficiently.


Your code is indeed straightforward. I can't think of an obvious way of using MatrixOP here (there are *only* about 170 MatrixOP functions)...

Here is another variation that does not use explicit loops:
// convert 1-layer image into a triplet wave
Function imageToXYZ(inWave)
    Wave inWave
   
    Variable rows=DimSize(inWave,0)
    Variable cols=DimSize(inWave,1)
    Variable nRows=rows*cols
    Make/O/N=(nRows,3) tripletWave
    tripletWave[][0]=mod(p,rows)
    tripletWave[][1]=trunc(p/rows)
    Duplicate/free inWave,ddd
    Redimension/N=(nRows) ddd
    tripletWave[][2]=ddd[p]
End


I would like to point out to the OP that they have to be careful in trying to use the Voronoi interpolation on data sampled on a rectangular grid. The problem is that the first step of the interpolation is the triangulation of the sampled data which is assumed to be sampled at random XY positions. As is obvious, the triangulation of samples on a rectangular grid is non-unique. This presents difficulties in the triangulation stage. IGOR attempts to introduce some perturbation to the XY locations in order to break the degeneracies.

Another consideration is that the triangulation step is performed in O(N^2) steps which is rarely justified when your data are already on a rectangular grid. So the bottom line is that you would have to have a very compelling reason to take this approach in the first place.

A.G.
WaveMetrics, Inc.

June 22, 2016 at 10:03 am - Permalink