<h1>Frequently Asked Questions</h1>
<p class="author">Last update: 1 February 2010</p>
I want to edit my image after blending. Can Enblend
create a layered output instead of a flat image, so I can adjust the
<p>No. A drawback of multiband
blending is that there is no good path for manually editing the
result. Enblend does not just find a single optimized blend mask
like the kind you are used to working with in Photoshop, where each
output pixel is a function of the input pixels directly underneath
it. Enblend looks at an entire region around a pixel in all of the
contributing input images and does an average. For typical digital
camera panos, these regions can be hundreds of pixels in diameter.
Furthermore, the size of these regions varies dynamically based on
how "edgy" the pixels are. Edges are blended across a much smaller
distance that open areas like the sky.</p>
I double-clicked on enblend.exe, and a window flashed briefly but nothing happened.
<p>Enblend is a command-line utility. You
have to run it from a DOS window and type parameters on the command
line. See Manual for more information on the available parameters.</p>
Enblend.exe only crashes on my Windows machine, even when I do simple things
such as asking for its version number.
<p>The default binaries of Enblend are compiled with
<a href="http://en.wikipedia.org/wiki/SSE2" target="_blank">SSE2</a> support.
If your CPU does not, there is a special version of the
that might work.</p>
Enblend says that images are redundant. What does this mean?
<p>It means that the image(s) that are
being added in the current blending iteration are completely
overlapped by the previous blended images. These images don't add
anything new to the panorama, so there is no way to draw a
transition zone between old pixels and new pixels. Also, Enblend
cannot decide for you that the new pixels are somehow "better" than
the old ones. It just keeps the old pixels and ignores the new ones.</p>
Why do my input images need alpha masks?
<p>Enblend is designed to work with
images with unusually shaped borders, like those that have been
warped by PanoTools. The alpha channel is required to indicate which
pixels in the file are part of the image versus the pixels that are
around the border of the image.<p>When the alpha channel is white
(255), it means that pixel should be part of the final panorama.
When the alpha channel is black (0), Enblend won't use this pixel
for blending. Don't use feathering in the alpha channel. You can
specify to use a pixel or not, but it does not mean anything to say
"use this pixel only N percent". That kind of constraint does not
work with the multiresolution spline technique.</p>
<p>You can use the alpha channel to remove unwanted pixels before
running Enblend. A common example is to erase pedestrians that are
moving between shots. If you erase the pedestrian in the alpha
channel, Enblend will consider this a "hole" in the panorama and
will fill in those pixels from a different image.</p>
What about multilayer TIFFs, like the ones that Nona produces?
<p>We need to extend the image
import/export capabilities of VIGRA to understand files that have
multiple images in them. For now, you can run tiffsplit on the
multilayer tiff and run Enblend on the resulting tiffs. You still
get the disk space savings in that each layer excludes the large
transparent borders. Tiffsplit comes with the libtiff-progs package.</p>
Why does the order of the images on the command line matter?
<p>Enblend works best when the overlap
between images is large. Therefore, you want to order the images
such that the next image overlaps the previous images as much as
possible.<p>Alternatively, try using the -a flag. This tells Enblend
to try to assemble as many non-overlapping images as possible before
blending. Sometimes this does a good job at maximizing the amount of
overlap, but it depends on your photos.</p>