I'm given large number of grayscale images, each as an array of grayscale values. For each image, I'd like to use the values as the alpha for drawing a single-color image of the same size in an arbitrary color. I can create a Bitmap where the RGB for each pixel is the solid color and the A is the corresponding grayscale value, and this works. However, I need to support multiple colors and creating a new Bitmap for each image/color combination seems terribly inefficient. Is there a more efficient approach in terms of memory and/or rendering speed?

Well, I answered my own question.

• Create an ARGB_8888 Bitmap where each pixel color is set to (gray << 24) | 0xFFFFFF.
• For each color, create a new PorterDuffColorFilter(color, PorterDuff.Mode.MULTIPLY).
• To render, create a Paint object and call setColorFilter() with the filter corresponding to the color to be used. Then call canvas.drawBitmap using the Bitmap and Paint objects.