Word Bounding Box Detection Using Gap Sizes

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Optical character recognition (OCR) algorithms face challenges in accurately recognizing text in images with multiple typefaces, arbitrary text placement, and varying text sizes, making it difficult to process complex text documents effectively.

Innovation Solution

A system that detects bounding boxes around words in an image by determining gap sizes, structuring element sizes, and using these to identify and correct line angles, allowing for efficient OCR by isolating regions with uniform backgrounds and similar font sizes, thereby improving recognition accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If OCR algorithms are applied to images with multiple typefaces, arbitrary text placement, and varying text sizes, then the system can process complex text documents, but recognition accuracy deteriorates

Engineering Contradiction:

Improveability to process complex text documentsVSAvoidrecognition accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the image processing task into distinct stages: detecting connected components, grouping them into lines, determining gap sizes, and creating bounding boxes. This segmentation allows each stage to handle specific aspects of the problem, improving overall accuracy while maintaining adaptability to complex layouts

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by determining gap sizes specifically for each line and using line-specific structuring elements for morphological operations. This localized approach allows the system to adapt to varying text characteristics in different regions of the document, maintaining high recognition accuracy across diverse text layouts

Inventive Principle:

Principle #3Local quality

2Measurement precision

If bounding boxes are determined for each word in complex documents, then OCR accuracy is improved, but processing complexity increases

Engineering Contradiction:

ImproveOCR accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary actions by detecting all connected components and grouping them into lines before determining bounding boxes. By preparing the data structure in advance with line groupings and gap size calculations, the subsequent bounding box determination becomes more straightforward and manageable, reducing processing complexity while maintaining accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces intermediate structures such as line groupings and gap size measurements that mediate between the raw image data and the final bounding boxes. These intermediaries simplify the complex task of bounding box determination by providing structured information about text layout and spacing

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS10127673B1Word bounding box detection

Publication Date: 2018.11.13 WORKDAY INC

AI summary

A system for determining bounding boxes includes the input interface and a processor. The input interface is configured to receive an image. The processor is configured to detect a line associated with connected components in the image; determine gap sizes within the line; determine a word structuring element size using the gap sizes; and determine bounding boxes for the line based at least in part on the word structuring element size.