good alternative is this inserted in a collection already?
The easiest thing to do is go to their archive with an ftp client like Fetch, grab the desired date range for the 4 years available while at lunch, set up a separate folder for the LARGE format, sort by size to pull out the many preliminary, defective, blank and wrongly masked files (proving nobody there ever looks at them), open earliest image to be used in gimp, 'open as layers' the rest, set the crop dimensions, note your upper left crop corner coords (to match for other years), set your rolling mask, capture 'new layer from visible', discard all but these, done.
There are a lot of variations on this (some posted, some not) such as darken only, weighted averaging, 12321 vertical convolutions, ramped palette squeezing from both ends etc etc. Thinking of these time series as tall columns of single fixed geo-pixels (at the center of a 3x3 patch of neighboring pixels) on which unexpected outliers are identified to which some sort of correction operation is applied.
No doubt some of these would improve markedly on the original set by taking down water vapor effects -- which transit more rapidly than ice moves/thins/compacts -- if we had some way of measuring how each method was doing, say rms error from actual values (defined at much higher resolution from S1A values), we could say what was the best approach at various points in the season for various cloud conditions. A lot of quantitative cloud data is collected each day but analysis at these forums has mostly just been ternary (cloud/thin cloud/no cloud).
As mentioned, I am offline traveling next 2-3 weeks. Keep up the great work everyone!
ftp://ftp-projects.zmaw.de/seaice/AMSR2/3.125km/