The binary derivative test: noise filter, crypto aid, and random-number seed selector

Random noise obscuring digit ized images or text can be removed by a new technique that recognizes the appearance o f randomness in short bit strings. This test makes uses binary derivatives. The derivative of a bit string is formed by recur sively XORing the bits pair- wise. The measure of random ness depends is the maximum range of the one's ratio (p) among various order deriva tives. Random pixel strings in digital images can be zeroed out leaving outlined regions where image details can be recovered by edge-detection or other tech niques. Binary derivatives can also assist cryptographers and cryptanalysts. And they can help simulationists avoid selecting random-number generator seeds that would bias their results.