Save an image with partial transparen​cy

Question

There appears to be no way to preserve alpha channel information when saving an image out of labview, even when using the IMAQ vision functions. That is to say, you can create a partially transparent image using labview, but you can't save it to disk. Is this correct?
global variables make robots angry

Answer 1

The issue boiled down to bad documentation. LabVIEW is cool,
but I really can’t wrap my head around how, given the very limited amount of
documentation, so much of it could contain goofy mistakes. The documentation
says that LabVIEW can’t handle a 32-bit image, and that it can only render
transparency as a binary mask.
Attached is a VI that demonstrates how to create partial
transparency in LabVIEW by building a 32-bit image with an alpha channel.
LabVIEW has no problem saving it out as a PNG. You can’t display it properly
with the picture indicator, but you can display it properly with the 3d picture
control. You have to ignore the documentation and switch the image bit depth
value to 32 and order the image data as groups of 4 bytes, alpha, red, green,
blue for each pixel.
Also, if you create a binary picture mask to be used with a
picture indictor, you need to add an extra column of bytes for no identifiable
reason. You’ll find that when LabVIEW reads in a PNG with an alpha channel, it
reads it in properly and creates a binary mask. You can save it out properly
and alpha channel info is preserved. The mask will be larger than expected
because of the weird extra column of bytes that isn’t mentioned in the
documentation. For instance, if you load a 600 x 400 PNG with an alpha channel,
the mask will be 30400 bytes long instead of 30000 bytes that would be expected
if the description of how the binary mask works was right.
Please note that there are two separate data structures (the
32-bit image data with an alpha channel and the binary mask with the weird
extra column) that are described incorrectly in the documentation for all of
the picture function help pages. Note that there are several lines in each affected
help page that need to be corrected.
I like LabVIEW because it’s great for communicating with NI
hardware, it’s easy to do parallel computing and it’s super easy to make a nice
GUI. I’ve used LabVIEW to write real-time target tracking software that works
on a wide range of targets in real video and I’ve used to it write a real-time
3D simulation using real imagery and real video including ground imagery,
transparent cloud imagery and real video of explosions. So, even though the
function libraries are not terribly robust, I know that LabVIEW has a good deal
of potential as a platform for serious programming work. I only started using
LabVIEW less than two years ago, so that should tell you that LabVIEW is fairly
intuitive to use.
However, I’m a little disappointed and frustrated with it
because I keep getting burned by poor documentation and silly bugs that eat up
programming time. LabVIEW is expensive software, and while the function
libraries are nice, they are somewhat limited. I think that National
Instruments should either offer much more functionality or do a better job
documenting and debugging the existing function libraries (of add-ons and
toolkits, too) in order to justify the cost of LabVIEW software.
Message Edited by Root Canal on 11-02-2009 05:33 PM
Message Edited by Root Canal on 11-02-2009 05:35 PM
Message Edited by Root Canal on 11-02-2009 05:37 PM
global variables make robots angry
Attachments:
transparency.zip ‏724 KB
nope.png ‏62 KB
32 bit with alpha channel small.png ‏599 KB