Hey Larry, No problem! Just wanted to make sure the email wasn't lost.
I ran a test with using ImageInput/Output with the config, and it worked as expected. I also tried reading through the ImageInput, and then did setpixel on an empty ImageBuf from the input, and that also worked as expected. So, I'm fairly confident that it was associating the alpha on read. There doesn't seem to be a config in the ImageBuf init for Python. I ran a test this morning, and no dice. So, that's the reason why I used the buf.reset(). I tested on oiio 1.5.19, and 1.5.22. I'd love to compile 1.6+, but don't have the time at work. :( Lastly, I've attached a couple of test files. The first one is not premultiplied, and the second one was ran through the example code: src: http://imgur.com/gF5grh5 dst: http://imgur.com/HrlIvNu Thanks for looking into this! - Scott On Mon, Mar 21, 2016, 11:47 PM Larry Gritz <[email protected]> wrote: > Sorry, was on vacation last week and still catching up on email. > > Is the issue that when you execute the code below, src.png and dst.png > don't match? Or are you confident that it is already wrong by the time it's > read in, even before the write? > > I think that > > config.attribute("oiio:UnassociatedAlpha", 1) > > > is the one you want to ensure that the read does not convert the > unassociated alpha. The attribute name searches are case-insensitive. > > buf = oiio.ImageBuf(src) > > # Read the file with the config > buf.reset(src, 0, 0, config) > > > That's a somewhat odd construct. Ick, it looks like Python lacks a > constructor that directly takes a config, like the C++ API has. I'll fix > that. That's why you're doing the reset(), right? > > In theory, the file being kept unassociated will also set > "oiio:UnassociatedAlpha" to 1 in the input spec (describing the data), > which is also used for the output spec (thus, providing the > "oiio:UnassociatedAlpha" = 1 to the ImageBuf.write. I'm not sure what's > going wrong. > > I'm worried about several issues: > > 1. If the 1.5 version you are using is too old or has bugs and this is > long since fixed. > 2. If the problem is on the read or when you write it back. > 3. Exactly how you know there is a problem (i.e., is there really a > problem). > > Perhaps you can clarify whether you think that the read values are wrong > in your Python script, or if it's only the written file that's wrong; > exactly what version of OIIO you are using (and better -- verify it it is > broken or working on the current 1.6 release and/or the current development > 'master'); and maybe you'd like to email me a png to try, just in case it's > something specific to one file, and I can try it on my end. > > -- lg > > > On Mar 17, 2016, at 6:08 PM, Scott Wilson <[email protected]> wrote: > > Hello, > > I am trying to open and save a PNG that contains RGB values that are not > associated with the A (or non-premultiplied) with OIIO and Python. I am > trying the following code to tell OIIO ImageBuf to read unassociated, but > it appears that it isn't respecting it. I'm restricted to OIIO 1.5.x due to > the only builds on Windows that I found with Python being from > www.lfd.uci.edu/~gohlke/pythonlibs/#openimageio. Am I doing something > wrong, or would I have to use ImageInput -> ImageBuf? If so, is there any > way to convert the ImageInput to ImageBuf without doing > ImageBuf.setpixel(x, y, z, pixel)? > > Thanks in advanced! > > # The code > import OpenImageIO as oiio > > if __name__ == '__main__': > src = "c:/temp/blah/src.png" > dst = "c:/temp/blah/dst.png" > > config = oiio.ImageSpec() > # Wasn't 100% sure what the attribute should be, so trying all possible > ones. :) > config.attribute("oiio:UnassociatedAlpha", 1) > config.attribute("oiio:unassociatedalpha", 1) > config.attribute("UnassociatedAlpha", 1) > config.attribute("unassociatedalpha", 1) > buf = oiio.ImageBuf(src) > # Read the file with the config > buf.reset(src, 0, 0, config) > # Write to a new png > buf.write(dst) > > _______________________________________________ > Oiio-dev mailing list > [email protected] > http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org > > > -- > Larry Gritz > [email protected] > > > _______________________________________________ > Oiio-dev mailing list > [email protected] > http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org >
_______________________________________________ Oiio-dev mailing list [email protected] http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
