Back to results
US 10,832,046 B1Granted
Systems and methods for processing document images
AssigneeArchitecture Technology Corporation
InventorsJafar Al-Gharaibeh, William Estey, Paul Nicotera
Life of this patent
Estimating the term…
Abstract
Disclosed herein are embodiments of an optical character recognition pre-processing software system, which is integrated into a language translation system to provide automated cleaning and correction of noisy and degraded document images to enable seamless and efficient optical character recognition processing and machine translation of information within the document images.
Claims18
A computer implemented method comprising:
- receiving, by a computer, a raw document image containing image noise and one or more distortions;
- determining, by the computer, a language associated with the raw document image;
- executing by the computer, a sliding window application configured to select a portion of the raw document image;
- transmitting, by the computer, the selected portion of the raw document image to a gating neural network hosted by the computer;
- selecting, by the gating neural network executed by the computer, a refiner neural network hosted by the computer based on one or more optical characteristics of the selected portion of the raw document image;
- pre-processing, by the selected refiner neural network executed by the computer, the selected portion of the raw document image to remove the image noise and the one or more distortion thereby generating a clean document image portion;
- stitching, by the computer, the clean document image portions with other clean document image portions to generate a clean document image;
- and
- transmitting, by the computer, the clean document image to a character recognition device such that the character recognition device recognizes one or more characters in the clean document image.
6 more elements
A system comprising:
- a non-transitory storage medium configured to store a gating neural network and a plurality of refiner neural networks;
- a processor coupled to the non-transitory storage medium and configured to:
- receive a raw document image containing image noise and one or more distortions;
- determine one or more languages associated with text in the raw document image;
- execute a sliding window application configured to select a portion of the raw document image;
- execute the gating neural network to:
- receive the selected portion of the raw document image from the computer;
- select a refiner neural network of the plurality of neural networks based on one or more optical characteristics of the selected portion of the raw document image, wherein the selected refiner neural network is configured to pre-process the selected portion of the raw document image to remove the image noise and the one or more distortion thereby generating a clean document image portion;
- stitch the clean document image portions with other clean document image portions to generate a clean document image;
- and
- transmit the clean document image to a character recognition device such that the character recognition device recognizes one or more characters in the clean document image.
8 more elements
Dependencies are inferred from the claim text.