Content-based document image classification
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for classifying one or more document images based on its content by determining blocks layout of the document image; recognizing the document image to obtain digital content data representing text content or the potential graphical content of the image; calculating feature values of the document image for features based on the digital content data and the blocks layout; and classifying the document image as belonging to one of document classes based on the calculated feature values.
1. A method for classifying a document image based on its content using a processor device, comprising:
accessing a set of features stored in memory;
analyzing the document image to determine blocks layout;
recognizing the document image to obtain digital content data representing text content or potential graphical content;
calculating, based on one or more features from the set of features accessed in the memory, feature values of the document image for the one or more features from the set of features, wherein the feature values are based on the digital content data and the blocks layout; and
classifying the document image as belonging to a document class from a set of document classes based on the calculated feature values.
2. The method of claim 1 , further comprising:
accessing the set of document classes and one or more sets of template images for at least one of the document classes;
analyzing the template images to determine block layouts;
recognizing the template images;
computing class feature values based on results of the analyzing of the template images and on results of the recognizing of the template images; and
storing the class feature values in the memory;
wherein the classifying the document image as belonging to the document class from the set of document classes is based at least in part on the class feature values.
3. The method of claim 1 , wherein the classifying the document image as belonging to the document class from the set of document classes further comprises calculating one or more confidence levels for the document image belonging to one or more classes from the set of document classes.
4. The method of claim 1 , further comprising:
performing preliminary processing of the document image to improve visual quality of the document image before the image analyzing and the image recognition steps.
5. The method of claim 1 , wherein classifying the document image as belonging to the document class from the set document classes further comprises:
calculating for at least one feature in the set of features a probability of the at least one feature being present in the one of the document classes; and
calculating overall probability of the document image belonging to the one of the document classes.
6. The method of claim 5 , wherein the probabilities of one or more particular features are weighted differently from other features.
7. The method of claim 5 , wherein classifying the document further includes storing the calculated probability of the at least one of the set of features in a classification database.
8. The method of claim 1 , wherein accessing the set of features includes accessing a library of at least one of:
relative spatial positioning of text content,
a frequency of word occurrence,
an amount of numbers in an image,
an amount of lexicalized words on a page,
an amount of present barcodes,
an amount of vertical/horizontal separators,
a ratio of objects in the page,
a vertical text orientation,
at least one key word, and
a location of an object within the image,
as at least one of the set of features.
9. The method of claim 1 , wherein classifying the document as belonging to the document class of the set of document classes further includes classifying the document into a graphical-only document, a text-only document, or a combination text and graphical document.
10. A system for classifying a document image based on its content comprising: a machine-readable storage device having instructions stored thereon; and data processing apparatus operable to execute the instructions to perform operations comprising:
accessing a set of features stored in memory;
analyzing the document image to determine blocks layout;
recognizing the document image to obtain digital content data representing text content or potential graphical content;
calculating, based on one or more features from the set of features accessed in the memory, feature values of the document image for the one or more features from the set of features, wherein the feature values are based on the digital content data and the blocks layout; and
classifying the document image as belonging to a document class from a set of document classes based on the calculated feature values.
11. The system of claim 10 , further comprising
accessing the set of document classes and one or more sets of template images for at least one of the document classes;
analyzing the template images to determine block layouts;
recognizing the template images;
computing class feature values based on results of the analyzing of the template images and on results of the recognizing of the template images; and
storing the class feature values in the memory;
wherein the classifying the document image as belonging to the document class from the set of document classes is based at least in part on the class feature values.
12. The system of claim 10 , wherein the classifying the document image as belonging to the document class from the set of document classes further comprises calculating one or more confidence levels for the document image belonging to one or more classes from the set of document classes.
13. The system of claim 10 , further comprising
performing preliminary processing of the document image to improve visual quality of the document image before the image analyzing and the image recognition steps.
14. The system of claim 10 , wherein classifying the document image as belonging to the document class from the set of document classes further comprises
calculating for at least one feature in the set of features a probability of the at least one feature being present in the one of the document classes; and
calculating overall probability of the document image belonging to the one of the document classes.
15. The system of claim 14 , wherein the probabilities of one or more particular features are weighted differently from other features.
16. The system of claim 14 , wherein classifying the document further includes storing the calculated probability of the at least one of the set of features in a classification database.
17. The system of claim 10 , wherein accessing the set of features includes accessing a library of at least one of:
relative spatial positioning of text content,
a frequency of word occurrence,
an amount of numbers in an image,
an amount of lexicalized words on a page,
an amount of present barcodes,
an amount of vertical/horizontal separators,
a ratio of objects in the page,
a vertical text orientation,
at least one key word, and
a location of an object within the image,
as at least one of the set of features.
18. The system of claim 10 , wherein classifying the document as belonging to the document class of the set of document classes further includes classifying the document into a graphical-only document, a text-only document, or a combination text and graphical document.
19. A non-transitory storage device having instructions for classifying a document image based on its content stored thereon that, when executed by data processing apparatus, cause the data processing apparatus to perform operations comprising:
accessing a set of features stored in memory;
analyzing the document image to determine blocks layout;
recognizing the document image to obtain digital content data representing text content or the potential graphical content;
calculating, based on one or more features from the set of features accessed in the memory, feature values of the document image for the one or more features from the set of features, wherein the feature values are based on the digital content data and the blocks layout; and
classifying the document image as belonging to a document class from a set of document classes based on the calculated feature values.
20. The non-transitory storage device of claim 19 , further comprising
accessing the set of document classes and one or more sets of template images for at least one of the document classes;
analyzing the template images to determine block layouts;
recognizing the template images;
computing class feature values based on results of the analyzing of the template images and on results of the recognizing of the template images; and
storing the class feature values in the memory;
wherein the classifying the document image as belonging to the document class from the set of document classes is based at least in part on the class feature values.
21. The non-transitory storage device of claim 19 , wherein the classifying the document image as belonging to the document class from the set of document classes further comprises calculating one or more confidence levels for the document image belonging to one or more classes from the set of document classes.
22. The non-transitory storage device of claim 19 , further comprising:
performing preliminary processing of the document image to improve visual quality of the document image before the image analyzing and the image recognition steps.
23. The non-transitory storage device of claim 19 , wherein classifying the document image as belonging to the document class from the set of document classes further comprises
calculating for at least one feature in the set of features a probability of the at least one feature being present in the one of the document classes; and
calculating overall probability of the document image belonging to the one of the document classes.
24. The non-transitory storage device of claim 23 , wherein the probabilities of one or more particular features are weighted differently from other features.
25. The non-transitory storage device of claim 23 , wherein classifying the document further includes storing the assessed probability of the at least one of the set of features in a classification database.
26. The non-transitory storage device of claim 19 , wherein accessing the set of features includes accessing a library of at least one of:
relative spatial positioning of text content,
a frequency of word occurrence,
an amount of numbers in an image,
an amount of lexicalized words on a page,
an amount of present barcodes,
an amount of vertical/horizontal separators,
a ratio of objects in the page,
a vertical text orientation,
at least one key word, and
a location of an object within the image,
as at least one of the set of features.
27. The non-transitory The storage device of claim 19 , wherein classifying the document as belonging to the document class of the set of document classes further includes classifying the document into a graphical-only document, a text-only document, or a combination text and graphical document.