Release v1.2.0. - Fix bug: The source image was modified (smoothed) if the binning was left to 1 and the gaussian prefilter value different from 0. - Added two new methods to detect reference surface: sparse max of mean and sparse max of std. They work the same as their non-sparse counterpart. However, we do not compute the filter value for every pixel of the source slice, but only a grid spaced by the size of the filter. And we multithread it. Greatly reduce omputation time for large images for which binning is not adequate. - Speed-up projection step for images that are stored in memory (as opposed to virtual stack that are streamed slice by slice from disk). - Update display (min & max range and time) during computation. So that the user is not presented with black images. Now we have a composite image, with min & max taken from the input image for the projection. For the reference image, we take the min & max from the max number of z-slices. There is still a bug where the displais is not fully refreshed on the first time-point, but this one I do not understand. - Fix current time-point not properly retrieved for preview. - Implement parameter persistence. The parameter values are remembered between Fiji sessions.