System of using neural network to distinguish text and picture in images and method thereof

• US 7,436,994 B2
• Filed: 06/17/2004
• Issued: 10/14/2008
• Est. Priority Date: 06/17/2004
• Status: Expired due to Fees

• Current Assignee: Primax Electronics Limited
• Original Assignee: Destiny Technology Corporation (Primax Electronics Limited)
• Inventors: Huang, Chun-Chia
• Primary Examiner(s): Bella, Matthew C.
• Assistant Examiner(s): Bayat, Ali
• Application Number:
  US10/868,977
• Publication Number:
  US 20050281455A1
• Time in Patent Office: 1,580 Days
• Field of Search: 382/156, 382/176, 382/186, 382/157, 382/151, 382/168, 382/195, 382/177, 715/863, 725/135, 380/54, 706/49
• US Class Current: 382/156
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

This specification discloses a system of using a neural network to distinguish text and pictures in an image and the method thereof. Using the knowledge of text recognition learned by the neural network in advance, images data of color brightness and gray levels in an image block are processed to generate a greatest text faith value. The system determines the text status of the image block by comparing a text threshold with the greatest text faith value. If the greatest text faith value is larger than the text threshold, then the image block is determined to contain text pixels; otherwise, the image block contains purely picture pixels. This achieves the goal of separating text and pictures in an image.

Claims

1. A system of using a neural network to distinguish text and pictures in an image, a set of training data being used to train the neural network in advance to generate text recognition knowledge, the system comprising:

an image block division module, which extracts gray-level image data of the image and divides the gray-level image data into a plurality of image blocks, each of which contains a plurality of block columns each of which is made of a plurality of continuous pixels;

a neural network module, which uses the text recognition knowledge to process the continuous pixels of the block column, generating a text faith value for each of the pixels and obtaining a greatest text faith value; and

a text determination module, which compares a text threshold with the greatest text faith value to determine the status of the image block, wherein the training data include photo-to-text data, white-to-text data, text-to-photo/white data, text-to-text data, no text data, data of text with more than one edge, and data of text with halftoning noise.

2. The system of claim 1, wherein the greatest text faith value is the maximum of the text faith values of the block column of the image block.

3. The system of claim 1, wherein the image block is determined to contain text pixels if the greatest text faith value is larger than the text threshold and to contain only picture pixels otherwise.

4. A method of using a neural network to distinguish text and pictures in an image, a set of training data being used to train the neural network in advance to generate text recognition knowledge, the system comprising:

extracting gray-level image data of an image and dividing the gray-level image data into a plurality of image blocks, each of which contains a plurality of block columns each of which is made of a plurality of continuous pixels;

inputting the gray levels of the continuous pixels of the block column, using the text recognition knowledge to process the continuous pixels to generate a text faith value of each of the block column and to obtain a greatest text faith value; and

determining the text status of the image block according to the result of comparing a text threshold and the greatest text faith value, wherein the training data include photo-to-text data, white-to-text data, text-to-photo/white data, text-to-text data, no text data, data of text with more than one edge, and data of text with halftoning noise.

5. The method of claim 4, wherein the greatest text faith value is the maximum of the text faith values of the block column of the image block.

6. The method of claim 4, wherein the image block is determined to contain text pixels if the greatest text faith value is larger than the text threshold and to contain only picture pixels otherwise.