Document image analysis method

Abstract

Digital images of documents such as bank checks are analyzed to determine acceptable image quality by capturing an image at a first or normal sensitivity and at a second or greater sensitivity. Both sets of image signals are subjected to speckle filtering to remove extraneous image elements or noise. The greater sensitivity signals are grouped in discrete areas having a predetermined number of pixels and black pixels which are connected are identified as an object in each discrete area. Each object is analyzed by measuring height, width, area, perimeter, aspect ratio, density ratio and perimeter to area ratio. Predetermined limits for these characteristics are applied to each object to classify the object as stroke-like or of another shape. Stroke-like objects are then compared with the object at the same position in the normal image field and a so-called area ratio of the number of black pixels in the normal image to the number of black pixels in the more intense image is determined. Stroke images having less than a minimum value of area ratio are subjected to a normal image missing pixel count and the image is deemed acceptable or unacceptable if a predetermined number of pixels are missing in the normal image. The method may be applied selectively to high rate document image capture and recording systems for verification of image quality and the health of the recording system.

Dependencies are inferred from the claim text.