Indexed PNG

I've extended the code of PNGImage.java to support indexed PNG files. I
needed this to be able to load map tiles from OpenStreetMap.

I've revised the constructor of PNGImage to take the necessary steps to
resolve the palette data with the helper methods of pngj.

Maybe you are interested in integrating this feature in future version
of JOGL. A patch is attached.

Thank you. I thought it was already supported.

Edit.: Your patch may require some more work.

On 03/01/2013 12:31 PM, ct [via jogamp] wrote:
> I've extended the code of PNGImage.java to support indexed PNG files. I
> needed this to be able to load map tiles from OpenStreetMap.
>
> I've revised the constructor of PNGImage to take the necessary steps to
> resolve the palette data with the helper methods of pngj.
>
> Maybe you are interested in integrating this feature in future version
> of JOGL. A patch is attached.
>

Thank you.

[Note: Patches w/o white space changes are better to read for differences]

I will review your patch and put it into consideration w/ Hernan's solution
  http://forum.jogamp.org/Able-to-disable-JOGL-s-PNG-decoder-tp4027701p4027739.html

~Sven

Sven, do you plan to use Hernan's solution? I can't remove all AWT dependencies from TUER because of this limitation :s

On 03/11/2013 10:29 PM, gouessej [via jogamp] wrote:
> Sven, do you plan to use Hernan's solution?
Yes .. and I will also consider 'ct's patch.

> I can't remove all AWT
> dependencies from TUER because of this limitation :s

No problem, will be fixed.

I am still testing a solution for Bug 665, after that I will do this one.

~Sven

Basically, my patch is based on a test case from PNGJ, as mentioned by Hernan.

The only thing one needs to do is to convert a scanline of palette indices to a scanline of colors. In my patch I've done exactly that by using PNGJ's helper classes. Further, the getPixelRGBA8 method now takes a scanline of int[], rather than of class ImageLine. This allows me to handle scanlines uniformly. I also cleaned up the PNGImage constructor a bit.

I have just modified your patch a little bit:
https://github.com/gouessej/jogl/commit/9fb623ef3c3feb30200c512e09b3507cccb76602

On 03/13/2013 09:58 PM, gouessej [via jogamp] wrote:
> I have just modified your patch a little bit:
> https://github.com/gouessej/jogl/commit/9fb623ef3c3feb30200c512e09b3507cccb76602

Thank you - it has no white space changes, hence enhance to read the diff.

So I assume CT's change is almost the same here - since I cannot diff
w/o having to do lots of work (pull merge and diff w/o white space change).

Looks good, similar to Hernan's code
  https://code.google.com/p/pngj/source/browse/pnjg/src/ar/com/hjg/pngj/test/SampleConvPalToTrueColor.java

Do we have a unit test ?

Could you extend 'TestPNGTextureFromFileNEWT', which currently tests:

  normal           test-ntscN01-160x90.png
  interlaced       test-ntscI01-160x90.png
  interlaced-gamma test-ntscIG01-160x90.png

and add for indexed:
  palette RGB      test-ntscP301-160x90.png
  palette RGBA     test-ntscP401-160x90.png

maybe we could also need (while we are at):

  normal RGBA       test-ntscN401-160x90.png
  interlaced RGBA   test-ntscI401-160x90.png

Please tell me if you like to do so, otherwise I will add the above
after the merge and test.

~Sven

I can do that in about ten hours. Yes, the fix I created by using ct's patch uses the same helper than Hernan's code.

This is in the pull request 60.

The unit test seems to fail, there is still something wrong.

On 03/15/2013 10:41 AM, gouessej [via jogamp] wrote:
> The unit test seems to fail, there is still something wrong.

Even though that is not good - great that the unit test catches it.

~Sven

I suspect it was already not working before modifying the code, I will give it another try.

On 03/15/2013 04:04 PM, gouessej [via jogamp] wrote:
> I suspect it was already not working before modifying the code, I will give it
> another try.

The unit tests are ofc working as they are (w/o indexed PNG),
dunno what you mean.

~Sven

The loading of test-ntscP301-160x90.png fails and it prevents all tests from working.

Edit.: I have forgotten to add the files :s

My latest commit fixes the remaining bug, it's ok now, you can test it again and merge :)