Abstract: A document imaging system includes means whereby a large number of anomalous condition indicators or flags that define anomalous conditions that may occur as a result of the operation of a document scanner and/or document processor. A computationally efficient subset of flags and flag combinations is defined, and these subsets of flags and flag combinations (system states) are related to a quantitative measure of the quality (degree of suspiciousness) of the related document images, the documents themselves, and units of work that contain a plurality of documents. The machine computation of quality is compared to a human perception of quality, and computational input parameters are adjusted to improve the match of computational quality to the human quality perception, thus teaching the machine to make proper determinations of image and document quality.