Use minimum of average template as its offset

[raacampbell]

See 19bfe8e

This stops weird patterns emerging on the image. We test for a while on dev before accepting fully.

[raacampbell]

The subtraction as implemented above will mean that different offset subtraction types are doing very similar things, if I understand correctly.

[raacampbell]

This is not working optimally. DG from Crick sends dataset AEW_A146 where the negative numbers are not present in the template we divide with but still there is some "brightness" in the darker corners. In the image below, the version with no background correction is on the right and on the left is the version with correction.

Part of the problem is low signal. But maybe we can do better?

[raacampbell]

Ok... It seems the problem is the change I made to the illuminationCorrector where I took this code:

 if userConfig.tile.doOffsetSubtraction
      m = stitchit.tools.getOffset(coords);
else
     m = 0;
end

and turned it into this:

if userConfig.tile.doOffsetSubtraction
        mO = min(aveTemplate.oddRows(:));
        mE = min(aveTemplate.evenRows(:));
        mP = min(aveTemplate.pooledRows(:));
else
        mO = 0;
        mE = 0;
        mP = 0;
end

Previously the template was seeing a subtraction that was controlled by the offsetType setting in the INI file. After this change, this ceased. It's true that this change helped with things on OG, but it has broken stuff elsewhere. The correct thing is likely to revert and alter the behavior of stitchit.tools.getOffset to handle OG, should this still be needed.

Specifically, the averageMinTile option is doing:

        aveTemplate = stitchit.tileload.loadBruteForceMeanAveFile(coords,userConfig);
        m=min(aveTemplate.pooledRows(:));
        if m>0
            m=0;
        end
        offset.(offsetType) = m;

So if the average is positive it doesn't subtract it. This maybe is not the right choice but this is where I should be looking for fixes. What I implemented previously was a hack.

[raacampbell]

Reverted the change and after bugfixes to the reversion: ba68071

But we still have issues with OG data. So we fix these specifically whilst not messing up the others...
EDIT: provisionally it looks like using averageTileMin is good enough for OG.

[raacampbell]

See Issue #145 - I noticed that the offset file is not updated when collate average images runs. We now delete the offset files whenever collateAverageImages is run. 16db7d3