With the multi-platform Android version, any given game background size need not be based on 640 x 480, but may take any rectangular size. In addition, the eventual playfields can be many sizes and usually will not not start at position 0, 0. Sprites will be scaled and their position on the screen scaled appropriately. Touch positions will be similarly displaced.
All collisions, zone effects, sprite movements and other screen events will be computed in the program on a game model which is at the basic background dimensions.
So, for drawing and detection of user touch, all positions, including background, sprite positions, dirty rectangle positions and touch positions, we have to make a transformation.
Rule 1. All images are pre-transformed to the correct scale. (This to save processing time during animation)
Rule 2. All position calculations are initially made on the base dimensions of the background playfield and proportionate sprite positions, and are only resized and displaced at draw time.
Rule 3 All touch co-ordinates are similarly transformed to base positions (by rescaling in reverse and subtracting the displacement) before being applied to the base model.
For example, the background. In any given scene, the background image does not vary. It is efficient to scale the image at the outset of the scene. Once scaled, and its displacement from position 0,0 determined, it can be drawn, in whole or in part, without having to scale the bitmap every time.
Similarly, sprite images will be resized, but all dimensions and positions will be reported to the base game model on the (in this case) 640x480 basis.
The actual placement and size of the sprite will be deternined in the draw stage.
There is a slight complication. Sprites can be scaled, flipped and mirrored , and this is done prior to the general scaling transformation.
It's not really so complex. Is it?