Token-Based Text Rendering for Scalable Digital Images

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Content providers face challenges in generating and storing digital images of text-based content due to inaccuracies in optical character recognition and high storage requirements, particularly when images need to be legible and scalable without excessive memory usage.

Innovation Solution

A method that converts scanned page images into token-based files by grouping similar tokens, creating vectorized tokens through morphological representation, and associating their positions, allowing for crisp and legible rendering at any resolution while minimizing storage needs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high resolution images are used to ensure legible text, then text legibility is improved, but storage requirements increase

Engineering Contradiction:
Improvetext legibilityVSAvoidstorage requirements
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The image is segmented into individual character tokens, which are then processed independently. Each character is extracted, recognized, and stored as a discrete unit with its position information, allowing selective rendering at high resolution only where text appears rather than storing the entire high-resolution image.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of storing the original high-resolution image, the patent creates a simplified representation by copying only the essential text information (character recognition results) along with position data. This allows reconstruction of legible text without preserving the full image data.

Inventive Principle:
Principle #26Copying

2Quantity of substance

If image resolution is reduced to minimize storage size, then storage requirements decrease, but text legibility deteriorates

Engineering Contradiction:
Improvestorage sizeVSAvoidtext legibility
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

By segmenting the image into individual character tokens and storing only their recognition results with position information, the patent avoids storing low-resolution image data while still enabling high-quality text rendering through selective reconstruction of character positions and forms.

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If manual correction is performed to fix OCR errors, then text accuracy is improved, but processing time and cost increase

Engineering Contradiction:
Improvetext accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the image processing into discrete character token recognition steps, allowing systematic handling of each character. This structured approach enables more accurate OCR processing while maintaining efficiency through automated token-by-token analysis rather than requiring manual review of entire pages.

Inventive Principle:
Principle #1Segmentation

4Adaptability or versatility

If scalable images are provided for various display sizes, then adaptability is improved, but storage requirements increase

Engineering Contradiction:
ImprovescalabilityVSAvoidstorage requirements
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent segments text into individual character tokens with stored position information, enabling flexible reassembly and scaling. Each character can be rendered at the appropriate size for different display resolutions without requiring multiple pre-scaled image versions, achieving scalability through programmatic reconstruction rather than storing multiple image variants.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP1999688B1Converting digital images containing text to token-based files for rendering
Publication Date: 2013.10.16 AMAZON TECH INC
  • EP1999688B1 patent drawingFigure 1
  • EP1999688B1 patent drawingFigure 2
  • EP1999688B1 patent drawingFigure 3

AI summary

A computer-implemented method is provided for converting a scanned-in electronic image into a token-based file. The method includes generally five steps. First, various tokens (i.e., graphical units) are identified in the electronic image. Second, the identified tokens having similar shapes are classified together to form a token group, to thereby form multiple token groups, each including one or more tokens having similar shapes. Third, in each token group, a representative token is found, which morphologically represents the shapes of tokens included in the group. Fourth, each representative token is converted into a vectorized token, which is a mathematical representation of the shape of the representative token. Fifth, each of the vectorized tokens is associated with the positions of the tokens in the electronic image represented by the vectorized token. Thus, upon rendering, the vectorized token is displayed to thereby create a page image consisting only of clean images of vectorized tokens.