A method of providing a portable true random number generator based on the microstructure and noise found in digital images is claimed and disclosed. Using the lowest significant bits of digitized images, strings of binary data are extracted. These raw strings are shown to pass the DIEHARD, NIST, and ENT tests for randomness for a robust selection of natural images. This information is available to, and may be processed by off-the-shelf technology including smartphones or other embedded devices without undue constraints on physical and environmental parameters. The method represents a significantly improved portable means of random number generation for all security, cryptographic, entertainment and PSI applications.