Report on enblend grey rim

[Klaus]

Hello,

I may have come closer to the mystery of the grey rim.

In Brief
The effect appears in the part of a second photo
processed by enblend that extends above the
first photo. It is related to the mask generation.
Using the --fine-mask option may be a workaround.

Disclaimer
My findings stem from one hour of investigation
looking at only two tiffs and the blending results.

Basis
Macmini Intel Core Duo 10.4.8 with enblend
cvs 20080120 packaged with MacPorts.

Method and Tools
Standard Preview image viewer as provided on Mac.
Gimp: loading two images as layers, then choosing
Difference as viewing mode, flatten image and
entering gamma=3 into Tools> Colour Tools>Levels.

Findings
The grey rim effect is dependent on the image order
(two image blend), affecting the second image and
there the area above the blending seam top position.
It is a gradual shade up to the top, mostly horizontal
in structure, although the may be an offset point.

It is the blending mask that matters. If one saves
the mask, inverts it and then loads it for blending
with reversed image order, there is little difference
between the blending results (small numerical stuff).

In particular, saving the mask from 2-1 (2 is the higher
image) and then using the inverted mask for a 1-2 run
does not give that grey rim.

Inspecting the TIFF as output by hugin, the image
edge and the alpha mask do align ok.

Setting the --fine-mask option, I was so far unable
to generate a grey-shade edge.

Conclusion
This effect is related to the mask generation.
It does not show up (limited stats) with --fine-mask.

Points not checked
Mask position and mask size in pixels,
are there differences between the runs?
Does a save-mask and and then load-mask
give the same result as the non-imported mask?
Reproduce effect with further images?


Can it be confirmed that the --fine-mask option
does not suffer from that bug?

Cheers

Klaus