I've applied the technique suggested by Knox-Thompson to allframes of a movie, and calculated the average phase. Thereconstructed result doesn't look like anything useful. Just using the average phase from the transform producessomething that looks like the average magnitude and zerophase reconstruction, but with some extra lobes at thecorners due to the phase bumps near DC. I decoupled thecorrelation calculation from the phase calculation, and thatproduced better results which might be a PSF candidate, butwhich is shifted slightly in the frame. The resulting phaseis smooth but has a slight gradient to it which isresponsible for the shift. To use as a PSF I just shift itso the maximum is at (31, 31).I wrote C code to do the forward multiplicativedeconvolution search, and applied that to several frames ofa movie, and then used the resulting images to get betterPSF estimates for each frame, and averaged them alltogether. The SWRI alphas are not online, so I've beenrunning everything on the 2 Macs I have here. I used the 3main PSF candidates (average magnitude and zero phase,average magnitude and correlated phase, and average PSF fromforward deconvolution frames) to compare their performancein iterated deconvolution (minimizing sum of squares error)of several movie frames. In all cases the 3rd PSF wasconsistently better than the other two, so I am runningforward deconvolutions on more movie frames to get more PSFsto average. I modified the multiplicative deconvolver to doadditive (and subtractive) perturbation of images, and theresult is that you get an image, of sorts, but it hasnegative values in it and isn't quite as good as thepositive definite image.I'm currently using random band-limited data with compactsupport to try to create an 'inverse PSF' to check out theclaim that deconvolutions can be turned into convolutions. If this works, it would mean that instead of spending a daydeconvolving each frame of data, we would spend a dayfinding an inverse PSF and then all deconvolutions of thatPSF would become (fast) convolutions with the inverse PSFinstead. I should know more about this next week.