You loose transperency in both 1st and 4th stage. In 1st because "Picture to Pixmap" stores transparency only in 4th byte of 32-bit image data, not in the mask. You have to manually convert this data to mask and perform your operations on the mask as well. But I've noticed that the transparency data generated by "Picture to Pixmap" in not compatibile to similar data generated by "Load PNG"... Bug? I have to do more testing on this. If you know something about the image you may generate mask yourself from color image data (i.e. using "Create Mask").
I would reccomend you using my BitMan library which helps with such situations and perform all operations on transparency automaticly.
Image rotation
in Machine Vision and Imaging
Posted
Thanks a lot. This was very helpful for me!