Hi Antonio,

Yes, the second part is still connected, and the symbol "files" has a value which is the list of file names. I reused the name "files" in the argument to the analyzeImageGroup function, which was a bit confusing. But appearing as an argument it is a formal parameter, and not the same. The function accepts a list of file names, which are still available as the value of the global symbol "files." But when used, analyzeImageGroup just expects a list of file names, each being a string. In the notebook, these are given without full paths because we have already set the default directory to the notebook directory, and the files are in it.

Here is some code which was executed after the notebook was executed. Notice that files still has a value which is our full list of file names. But I can also give the function a list of file names I have typed in. And I can also Map the function onto a list of lists, where each bottom level list is a list of file names. Now I get a list of analysis results, one result for each of the bottom level list of file names. This could be used to process an entire set of data, consisting of separate lists for each group of images which should be added and analyzed.

In[63]:= files

Out[63]= {"fiber3_0.fits", "fiber3_1.fits", "fiber3_2.fits", \
"fiber3_3.fits", "fiber3_4.fits"}

In[64]:= analyzeImageGroup[files]

Out[64]= {{476.983, 500.031}, 386.469, {"fiber3_0.fits", 
  "fiber3_1.fits", "fiber3_2.fits", "fiber3_3.fits", "fiber3_4.fits"}}

In[65]:= (* Here we explicitly give it a list of file names *)

In[66]:= analyzeImageGroup[{"fiber3_1.fits", "fiber3_2.fits"}]

During evaluation of In[66]:= InterpolatingFunction::dmval: Input value {0.652631,498.308} lies outside the range of data in the interpolating function. Extrapolation will be used. >>

During evaluation of In[66]:= InterpolatingFunction::dmval: Input value {0.176204,498.308} lies outside the range of data in the interpolating function. Extrapolation will be used. >>

During evaluation of In[66]:= InterpolatingFunction::dmval: Input value {-0.300222,498.308} lies outside the range of data in the interpolating function. Extrapolation will be used. >>

During evaluation of In[66]:= General::stop: Further output of InterpolatingFunction::dmval will be suppressed during this calculation. >>

Out[66]= {{471.156, 498.285}, 387.422, {"fiber3_1.fits", 
  "fiber3_2.fits"}}

In[67]:= (* The errors are not a problem, but can be turned off if we \
want *)

In[68]:= Off[InterpolatingFunction::dmval]

In[69]:= analyzeImageGroup[{"fiber3_1.fits", "fiber3_2.fits"}]

Out[69]= {{471.156, 498.285}, 387.422, {"fiber3_1.fits", 
  "fiber3_2.fits"}}

(* We can Map the function onto a list of lists to get a list of \
results *)

In[70]:= listOfLists = {{"fiber3_1.fits", 
    "fiber3_2.fits"}, {"fiber3_3.fits", "fiber3_4.fits"}};

In[71]:= analyzeImageGroup /@ listOfLists

Out[71]= {{{471.156, 498.285}, 
  387.422, {"fiber3_1.fits", "fiber3_2.fits"}}, {{473.682, 500.875}, 
  386.866, {"fiber3_3.fits", "fiber3_4.fits"}}}

(* Turn the error message back on *)

Hi Antonio,

Just about the first line in the notebook uses FileNames function to generate a list of the file names as strings and assign the list to the symbol files. But any method that generates such a list would work the same. You could also Map the analysis function onto a list of lists to analyze severa groups of files.

Yes. I would love to see the paper. You could attach it or send it to my email.

Best, David

Hi Antonio,

I think convergence will be influence by the image data. With sufficiently noisy the process could just wander around. A choice of a center produces a set of maxima, which produces a different center, etc. There could be patterns that even cycle without convergence. For sufficiently noisy image data we might be just wandering until we happen to get 2 centers in a row which are sufficiently close. But the process really doesn't converge in a continuous way. It might converge better for smoothed data, but then you have to ask if that particular converged result is really better than one of the results from acting on non-smoothed data.

I was looking at a different calculation which might be interesting. I calculate the set of maxima around the ring as before. But instead of calculating a center from the XY positions, I use the XYZ positions and fit to a plane. The Z is the image intensity. It is interesting the way the ring is sharp and intense toward one side, and broader and less intense toward the other. One possible explanation is that it is as an object more uniform, but in the imaging system it is tilted with respect to the optical axis -- so the broader less intense side is really just out of focus.

This "tilt" is of course just a parameter. The X and Y is in units of pixels (which are distance), and the Z is in intensity, which is in different units, so ArcTan of intensity divided by distance is a rather strange thing.

Here is a code snippet. It calculates a tilt of almost 1 degree for the plane fitting the maxima:

In[33]:= fit = LinearModelFit[centeredMaxima, {x, y}, {x, y}];

In[34]:= parameters = fit["BestFitParameters"]

Out[34]= {-2.91103*10^-17, 0.00012698, -0.00154588}

In[35]:= tilt = Norm[parameters[[{2, 3}]]]/Degree

Out[35]= 0.0888707

In[36]:= fitFunction = fit["BestFit"]

Out[36]= -2.91103*10^-17 + 0.00012698 x - 0.00154588 y

Here is a set of maxima and the best fit plane:

Hi David, I need more time to evaluate your latest code. So far I could take apart and understand your previous code almost completely. I have not a final conclusion yet. I can say that the code seems to work as it should and that their algorithm seems ok. However, I'm not too sure about the result, I need to continue investigating. Give me some more days......

This experiment is not confidential, You can ask what you want. This is a situation in which light from a laser is incident on the input fiber with an angle to the surface. The result is that the output light is a ring-shaped projection. The image that you see, is obtained with the help of a beam spliter reflecting the ring projected onto a target at about 1 meter from the output of the fiber. The analysis of the results can give us informations about the polishing angle of the fiber in test.

Thanks a lot.....

Best regards,

Antonio

Hi Antonio,

The code I posted yesterday uses an algorithm like you first described -- it determines a point "centered" with respect to a collection of maxima along radials, and it iterates. It may work.

Regarding the addition of a list of images, that is what it does. You give it a list of file names. It imports those files as images, adds them to make one image, and analyzes it. This can be immediately extended as you describe above. If you partition the file names into a list of lists, where each low level list is a list of files you want added and analyzed, then you will get a list of results. Look at the function named Map. Map[f,{a,b,c}] outputs { f[a], f[b], f[c] }.

So if you have a large list of files which you want to be analyzed in groups, you partition them:

partitionedFiles = { {file1, file2, file3, file4, file5}, {file6, files6, file8, file9, file10} }

And then

Map[ AnalyzeImageGroup, partitionedFiles]  or the shortcut AnalyzeImageGroup/@partitionedFiles

will produce an analysis for each of the sublists as a group. You will get a list of results, one result for each group. And an annotated graph will be saved as a png for each of the groups, using the name of the first file in the group for the name of the png.

Kind regards,

David

David, I'll take some time to fully understand and test your code by myself. I can see clearly que your approach is different of mine.This is something really very liberating.

As I said before, the code needs to repeat the calculation process for 20 pictures or more, always adding each group of five images in sequence. Can you do this repetition?

My big difficulty with Mathematica is to imagine repetitive processes in long codes without use of looping.

Thanks

Cesar

Dear Frank, Thanks for your suggestion. I have a tendency to use looping because my only computer training was over 30 years with FORTRAN and looping was a very reasonable option. I will work hard with your suggestion ......... Thanks