Back to results
US 8,554,021 B2Granted
Finding similar content in a mixed collection of presentation and rich document content using two-dimensional visual fingerprints
AssigneePalo Alto Research Center Incorporated
InventorDoron Kletter
Life of this patent
Estimating the term…
Abstract
Visual fingerprinting is used to provide a robust and highly effective method of finding similar content in a large document collection of rich document content composed of multiple text, line-art, and photo image objects. The visual fingerprints capture unique two-dimensional localized aspects of document appearance. The visual fingerprints are highly distinctive; fast for lookup; compact for storage requirements; and scalable to large document collections.
Claims25
A method of detecting document image content, the method comprising:
- separating a document image page into a plurality of image layers, each of the image layers having distinct characteristics from other ones of the image layers each corresponding to distinct category types;
- generating two dimensional visual fingerprints of the plurality of image layers, wherein distinct two dimensional visual fingerprinting techniques are used for each of corresponding ones of the distinct image layers;
- labeling and storing the formed two-dimensional visual fingerprints of the image layers by the category types;
- repeating the separating of the document image page, generating the two dimensional visual fingerprints, and labeling and storing the two dimensional visual fingerprints for a plurality of document image pages;
- and
- detecting image content between a query document image page and the plurality of document image pages, using the labeled two dimensional visual fingerprints, wherein the method is performed by use of at least an electronic processor.
3 more elements
A method of classifying pixels of an image, fingerprinting the image, and using the fingerprints of the image for content detection, the method comprising:
- defining a local image window;
- examining image properties of the pixels of the image in the local image window centered at a current pixel of interest by:
- (i) analyzing with a color measure detector color distribution of the pixels in the local window, to determine the number and modality of colors,
- (ii) analyzing with a second contrast measure detector normalized contrast and edge sharpness for edge pixels within the local window, and
- (iii) analyzing with a third texture measure detector normalized texture response for the pixels within the local window;
- labeling and storing pixels outputted from the detectors as at least one of a text pixel, a line art pixel and a photo image pixel in a decision lookup table;
- copying content of the pixels to one or more of a text image layer, line-art image layer and photo image layer in accordance with the labels applied to the pixels;
- forming two-dimensional visual fingerprints of the text image layer, the line-art image layer and the photo image layer, the forming of the two-dimensional visual fingerprints of each of the image layers achieved independently of the other image layers;
- storing the formed two-dimensional visual fingerprints of each of the image output layers;
- and
- comparing the stored two-dimensional visual fingerprints to determine content of a query document image page that matches the stored two-dimensional visual fingerprints, wherein the method is performed by use of at least an electronic processor.
8 more elements
A system of detecting document image content, the system comprising:
- an electronic processor configured to:
- separate a document image page into a plurality of image layers, each of the image layers having distinct characteristics from other ones of the image layers;
- generate two dimensional visual fingerprints of the plurality of image layers, wherein distinct two dimensional visual fingerprinting techniques are used for each of corresponding ones of the distinct image layers;
- store the formed two-dimensional visual fingerprints of the image layers;
- repeat the separation of the document image page, generation of the two dimensional visual fingerprints, and store the two dimensional fingerprints for a plurality of document image pages;
- select a query document image page;
- separate the query document image page into a plurality of query image layers, each of the image layers having distinct characteristics from other ones of the image layers;
- generate two dimensional visual fingerprints of the plurality of query image layers, wherein distinct two dimensional fingerprinting techniques are used for each of corresponding ones of the distinct image layers;
- compare the two dimensional visual fingerprints of the plurality of query image layers to the two-dimensional visual fingerprints of the plurality of image layers;
- and
- determine content of at least one of the plurality of document image pages and the query image page.
8 more elements
Dependencies are inferred from the claim text.