Can I use the GPU to speed gaussian fit on a large array of vectors?

Hi all, I'm trying to write a script that will analyze a group of video frames on a pixel-by-pixel basis. So, I have 200 640x480 images. I pull them in so I have one big 640x480x200 matrix. Thus far I have been using nested loops (and parfor) to analyze each 200 element vector in the 640x480 array. I need to do a Gaussian fit on the vector and return the peak of the Gaussian fit (coefficient b1). My current problem is that it takes a long time to run (over 1 hr) and I'm only able to halve the time using parfor on a dual core CPU.

So my questions are: Can I use the GPU to do more operations in parallel? The fit function is not supported for GPU. Can I use it in arrayfun? Do you have any other tips that would speed this calculation?

I've done some timing and the fit function is the slowest part of the script. I tried to speed it up by limiting iterations and function evaluation. But mostly it's repeating the calculation 307200 times that is taking so long. Eventually, I need to use this script to analyze data with more than 5000 frames so any speed would help greatly.

Here's my code:

if true
%FA is a 640x480x200 double
for j = 1:640
  parfor k = 1:480
      data=squeeze(-FA(j,k,:)); %extracts z-vector from single pixel
      [pks,locs] = findpeaks(data); %smooths peaks over interference fringe modulation
      pks = pks + abs(min(pks)); %moves baseline to zero
      [mx, mxloc] = max(pks); %IDs location of maximum
      f = fit(locs, pks,'gauss1','MaxIter', 1, 'MaxFunEvals', 3, 'Exclude', locs < mxloc - 5 & locs > mxloc + 5);
      fitcoeff = coeffvalues(f);
      CellArray(j,k) = fitcoeff(1,2); %returns b1 fit coefficient to array
  end 
end
end

Thanks in advance!