Do the images have large areas of flat colours? Are they weird shapes? You might be able to split each image into many smaller tiles, and find that many of the tiles are empty, or are filled with just one colour. Then you don't need to provide high-resolution images for those tiles. You could automate the process of cutting up the images.
Alternatively, maybe you can look at performing PNG or JPEG decompression on the fly, so you can use these far more compressible image formats. If your images are very simple, (e.g. flat colour vector graphics), even simple RLE encoding could reduce the size a lot without the complexity of finding or implementing a PNG or JPEG decompressor.
Also if they are vector graphics then you may be able to re-render them on the fly from the vector sources, which will let you match whatever the device's native resolution is.
↧
