I'm not sure it's as difficult as you suggest. As long as you don't let applications touch the pixel data of decoded images, it's all within libraries, and you can change the bit depth of buffers and change how they map to physical pixels easily.
Most frameworks don't let applications touch pixel data without jumping through some hoops, because by restricting it you can implement things like lazy loading, GPU jpeg decoding GPU resizing, etc.
Most frameworks don't let applications touch pixel data without jumping through some hoops, because by restricting it you can implement things like lazy loading, GPU jpeg decoding GPU resizing, etc.