Key-Point Text Localization for Training-Free OCR

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing text localization algorithms in OCR systems face challenges in efficiently identifying and isolating regions of interest in scanned documents, particularly for handwritten text, requiring extensive training and re-training for different sources and writing styles.

Innovation Solution

A key-point based text region identification algorithm that combines ROI identification, space identification, and ROI clustering to accurately isolate words without the need for training, using pre-processing, ROI identification, space identification, and ROI clustering steps to generate bounding boxes around words.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional text localization algorithms are used, then text identification can be performed, but extensive training and re-training are required for different sources and writing styles

Engineering Contradiction:

Improveadaptability to different writing stylesVSAvoidtraining time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs self-calibration by automatically detecting key points and computing transformation matrices without requiring external training data or manual adjustment. The algorithm adapts to different writing styles and document types autonomously through iterative refinement of detected text regions

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The method dynamically adjusts detection parameters based on the input document characteristics. By computing transformation matrices and refining bounding boxes based on detected key points and text regions, the system adapts its parameters to match different writing styles and document formats without re-training

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If traditional text localization algorithms are used, then text regions can be identified, but the accuracy and efficiency of word isolation is insufficient

Engineering Contradiction:

Improveaccuracy of word isolationVSAvoidefficiency of text identification

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The algorithm segments the document into discrete text regions by detecting individual key points and computing bounding boxes for each word or character. This segmentation approach enables precise isolation of individual words while maintaining processing efficiency through automated detection

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system replaces manual text localization mechanisms with automated computer vision algorithms. By using key point detection, transformation matrix computation, and automated bounding box generation, the system achieves both high accuracy and efficiency without human intervention

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250329180A1Key-point based text region identification

Publication Date: 2025.10.23 OPEN TEXT CORPORATION

AI summary

Systems and methods for text localization are provided. Various embodiments of the present technology provide systems and methods for improved text localization algorithms that will help in enhancing the efficiency of text identification algorithms used for recognizing text in scanned documents prior to performing OCR, or other related applications. In some embodiments, regions of interest are identified on an image document indicating locations on the image document where text may be present. Individual words in the image document are identified based on space identification and region of interest clustering algorithms applied to the regions of interest in the image document.