Add alpha channel to a BufferedImage

Question

Hi, I am making a small jigsaw puzzle type game. I read an image from a file into a BufferedImage using ImageIO.read(). From this image I cut out rectangular pieces using getSubimage(). Then I would like to mask out the small figure cuts that should be along the edges of each piece. I'm thinking that I would like to do this by alpha transparency - making the background show through the irregular edges. So, my problem is that the images returned by ImageIO.read() does not necesarily have an alpha channel.
Is there a simple way to add an alpha channel to an existing BufferedImage?
Or does anyone have a better solution?
P.S. Currently I intend to use a single circle at each edge, either inside the piece as a hole or filled outside the edge. These circles can easily be drawn, but I plan to eventually expand this to be able to make arbitrarily shaped pieces.

Answer 1

Is there a simple way to add an alpha channel to an
existing BufferedImage?Not for the general case. For example, I used ImageIO.read to load in a jpeg image. It's 50x60 pixels, and as an image it has 3 channels: red, green and blue -- no alpha. The pixel data is stored in a DataBufferByte, which has a single array of 9,000 bytes (3 channels x W x H). Given the packed structure of that data, there is no efficient way to slide in alpha values -- there's no room.
One thing you can always do is copy the data from this BufferedImage into another BufferedImage that has an alpha channel. Here's some quick and dirty code to do it:
static BufferedImage addAlpha(BufferedImage bi) {
     int w = bi.getWidth();
     int h = bi.getHeight();
     int type = BufferedImage.TYPE_INT_ARGB;
     BufferedImage bia = new BufferedImage(w,h,type);
     Graphics2D g = bia.createGraphics();
     ImageObserver imo = null;
     g.drawImage(bi, 0, 0, imo);
     g.dispose();
     return bia;
Or does anyone have a better solution?
P.S. Currently I intend to use a single circle at each
edge, either inside the piece as a hole or filled
outside the edge. These circles can easily be drawn,
but I plan to eventually expand this to be able to
make arbitrarily shaped pieces.1. The code given above is sufficient to take the original image and create the same image, but with an alpha channel. Don't worry about the fact that you're copying pixel data. Are you generating 100 jigsaw puzzles per second?
2. Another possible solution is to start with the entire image, perhaps with no alpha channel, and draw individual jigsaw pieces by adjusting the Clip property before drawing each piece. Now, this may be a little slow, because calculating a fancy clipping shape with half-circles etc... may complicate the rendering
3. My Conclusion. Because a jigsaw application would involve lots of repainting, rotating pieces and dragging them, I'd do the following:
a. Create buffered images (with alpha) for each piece. I'm not using subimages because I'm trading more space for faster redrawing time. To be rigorous, one should do it both ways and compare the results.
b. Just once, for each of these piece images, using a clipping shape, draw from the original image (with or without alpha) onto the piece image to generate the image of one piece. If your jigsaw game allows pieces to be rotated 90/180/270 degrees you may want to pre-generate those rotated piece images to speed up drawing the pieces in rotated orientations. Again, one should do it both ways and compare the results.
--Nax