| 摘要 |
A method and system for determining whether a digitized page is intended to be viewed in a first, portrait orientation or in a second, landscape orientation that is rotated with respect to the first orientation. The method includes a first step of examining an image of the digitized page on an image scanline by image scanline basis to detect occurences of graphic elements having heigths, widths, and aspect ratios that are within predetermined limits. The range-defining thresholds are determined to correspond to the expected dimensions of "charactersized" graphic objects. For each detected occurrence, the method determines a spatial relationship of the detected occurrence to other adjacently disposed objects. The method further operates, in accordance with a plurality of the determined spatial relationships, to declare the digitized page to have one of the first format and the second format. The step of determining operates to ascertain an amount of a spatial separation between the detected occurrence and an adjacent object on the same scanline and to an adjacent object on a succeeding scanline. The step of examining includes a step of forming, for each scanline, a list of spans of image pixels each having a predetermined value, and the step of determining a spatial relationship determines, for each span of the list, a distance to a next span on the same scanline, and for a last span of an image object, a distance to a first open span on a succeeding scanline. Each image object has an associated bounding box and a clear pedestal having a width that is a function of the width of the bounding box. The distance to the first span on a succeeding scanline defines a height of the pedestal. |
