Ah.. I see you edited your posts while I was replying...
---------------------
1920x1080 images in a few KB, eh? What kind of image data are we talking here? Woodland scene photography, or simple vector-style graphics? Some lossless algorithms can squeeze some images into a very small size, while others they struggle with (.png). Some algorithms do well with complex data (.jpg), but offer few other benefits (e.g, lossy, ineffective on simple data, etc.) Is it lossy? Is it most efficient at complex data or simple? Does it support alpha?
Interesting claims you have made here. If it is as good as you claim, you should patent it asap. However, to be a Johnny Raincloud, lots of research has been going into image (and data more generally) compression techniques for decades now.........