Method of identifying the script of a document irrespective of orientation - 6005986









The present invention is a method of identifying the script and orientation of a document image by identifying each set of connected pixels in the document image; computing the number of pixels in each set of connected pixels; computing the horizontal mean position of the pixels; computing the vertical mean position of the pixels; computing the horizontal extent of the pixels; computing the vertical extent of the pixels; computing a plurality of moment values for each set of connected pixels in the document image using a unique normalized centered moment calculation; grouping the moment values according to moment type; sorting the moment values within each moment group according to moment value; selecting moment values from each rank ordered moment group in order to characterize the document image; comparing the selected moment values to moment values for representative document images in a number of scripts and orientations; defining the script and orientation of the document image as being the same as the representative document that it compares most favorably to; selecting at least one script of interest to the user; rotating the document image to a user-defined orientation if the document image is of a script of interest to the user; and presenting the rotated document to the user if the document image is of a script of interest to the user.Method of identifying the script of a document irrespective of orientation600598621/12/199903/12/19971999748Ratner; Alan S.US Patent and Trademark OfficeGoogle Patent Searchpatentimages.storage.googleapis.com/pages/US6005986-1.pngUniquedocument imagemoment valuesscriptpixelsconnected pixelsorientationusermoment groupsetinterest