Image analysis provides a fast, inexpensive, non-invasive way to rank samples according to their potential value scientifica useful feature in a data-rich environment. We describe a technique to classify images of rock samples as targets of interest with respect to biogenicity: are the samples likely to contain evidence for life, or not? Using a data cluster model constructed from terrestrial rock samples with known properties, we analyzed the potential biogenicity of stromatolites found on Earth and several rocks imaged by the Mars Exploration Rovers. Twelve properties directly measured from the digital images (gzip and png compression ratios, entropy, energy ratio, and 8 Gray-Level Co-occurrence Matrix features) were used in the data cluster model based on a suite of known rocks and minerals. Two macroscopically similar Earth stromatolites (laminated columnar branching structures), one of known biogenic origin and the other of known abiotic origin, were classified using the cluster analysis. All but one of the Mars samples were classified as abiotic. One sample (Humphrey) was flagged as meriting further observation and analysis; this rock was previously identified as containing evidence of ancient flowing water. While conclusive judgments about biogenicity are unlikely to be made solely on the basis of image features, this analysis can provide first estimate of the importance of a follow-up search for other biosignatures (e.g., isotopic or chemical analysis). Given a ranked list of potential targets, more ex pensive analyses, in terms of integration time or consumables, can be focused on targets with maximal potential interest.