Document image processor with defect detection

Abstract

Automatic detection of defective pixel locations from digital images of scanned documents is provided. A list of tentative defect locations is kept, and as each document is scanned, entries are added to the list. As subsequent documents occur with the same color pixel in the same location, a count for that tentative defect location is incremented, but if the color of the pixel in that location changes, the count is reduced or zeroed. If a count for a location is incremented above a threshold, the tentative defect is flagged as an actual defect and a defect detector outputs or flags that pixel location as being an actual defect location. If memory size is constrained, the number of entries in the defect list is capped at some maximum size and a tentative defect location is added to the list only if an entry is available. To avoid the bias of the defect list being filled with pixel locations from only the first portion of documents scanned if the maximum number of entries is much less than the number of tentative defect locations, cursors changeably identify preferred regions of a document and tentative defects are only accepted for placement on the defect list from the preferred region.