OddThinking

Lossy compression algorithms take advantage of the way humans sense sound (in the case of MP3), colours (in the case of JPEG) and movement (in the case of MPEG) in order to remove unnecessary detail and to thus take up less storage space.

I am working on a new lossy compression system for images that takes advantages of two key facts about photographs:

1. A large number of photos taken are of babies.
2. All babies look the same.

The compression system will detect that the image contains one or more babies, and will replace the image data with a simple integer flag counting the number of babies in the photo. (It is anticipated that a single nybble (4 bits, range 0 to 15) should be sufficient for most applications.)

The decompression system will superimpose on the basic image n copies of an image of a canonical baby. Note: The codec standard does not dictate the exact bitmap – any baby-equivalent image may be used.

Initial modelling suggests that the entire works of Anne Geddes could be compressed onto a single floppy disk. Early trials suggest that the quality-loss is undetectable to members of our first test group. (We hope to repeat the trials with larger, more diverse consumer groups in the future, perhaps including some demographics outside of the “childless male” group we started with.)

Extensions to the format are expected in the future – including ultrasound images which, according to our Information Theory calculations, can be completely replaced with a simple white-noise generator.