A recording technique has been developed that allows machine-readable digital information to be encoded within human-readable imagery. Each picture element in the recorded image consists of a square matrix of discrete spots, each spot taking one of several possible optical densities. A specific format consisting of a 2 x 2 spot matrix, with 8 quantization levels per spot, has been investigated in detail. This 2 x 2 x (8) format encodes 12 bits of machine-readable digital information per pixel and can exhibit 330 discrete readable pixel densities. This technique can be used, for example, to record human-readable imagery derived from EO sensors with 10 bits of dynamic range. The full sensor data base could be recovered from the imagery, by use of a suitable reader, without loss of digital information. In addition, collateral information such as height data, annotation, etc., can be encoded within the digital data in machine-retrievable format in a way that does not affect the human-readable information. This paper discusses algorithms developed for mapping the digital data into human-readable pixel densities, some techniques for error correction and pixel identification, and examples of collateral data fusion. Results of a digital data retrieval experiment are presented.