Query-Guided Image Text Prediction for Targeted OCR Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing optical character recognition (OCR) systems face challenges in accurately detecting and recognizing text in real-world images due to unpredictable text locations, noise, and the need to extract specific elements of interest rather than all text, especially in complex scenes like streets or documents.

Innovation Solution

A system that integrates text detection and character recognition in a single end-to-end process, using feature extraction and decoder modules guided by queries to predict specific text elements of interest, employing neural networks for orientation-invariant text recognition and refining outputs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If traditional image processing techniques are used to extract text, then the process requires manual intervention and is time-consuming, but automation reduces productivity and increases errors

Engineering Contradiction:
Improvetext extraction automationVSAvoidtext extraction speed
Core Design Contradiction:
Extent of automationVSProductivity

Solution Approach 1:

The system enables self-service by automatically detecting, segmenting, and recognizing text in images without requiring manual intervention. The automated text extraction system processes images independently, eliminating the need for human operators to manually transcribe or transcribe text, thereby achieving both automation and high productivity simultaneously

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces manual mechanical text extraction processes with an automated computational system. Instead of human operators physically reading and transcribing text, the system uses image processing algorithms, optical character recognition (OCR), and machine learning models to automatically extract text information, substituting mechanical human labor with automated digital processing

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual text extraction is performed, then accuracy can be maintained, but time consumption increases and productivity decreases

Engineering Contradiction:
Improvetext extraction accuracyVSAvoidtext extraction time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-processing images to enhance text visibility and clarity before extraction. This includes image normalization, contrast enhancement, and noise reduction techniques that prepare the text data in advance, ensuring high extraction accuracy while reducing the time required for actual text recognition and transcription

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The automated system incorporates feedback mechanisms where the extracted text is validated and corrected automatically. The system can compare extracted text against expected patterns, detect errors, and self-correct, maintaining high accuracy levels while processing images much faster than manual extraction, thereby resolving the contradiction between accuracy and time consumption

Inventive Principle:
Principle #23Feedback

3Productivity

If automated text extraction systems are implemented, then productivity increases, but system complexity and computational resources are required

Engineering Contradiction:
Improvetext extraction speedVSAvoidtext extraction system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system applies segmentation by dividing the complex text extraction process into distinct modular stages: image preprocessing, text detection, character segmentation, OCR recognition, and post-processing. This segmentation allows each component to be optimized independently and simplifies the overall system architecture, making the high-productivity automated system more manageable and less complex

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a universal text extraction system that can handle multiple image types, languages, and text formats within a single integrated framework. The system uses multi-functional algorithms that can process various document types (scan, photograph, handwritten text) and languages simultaneously, reducing the need for multiple specialized systems and thereby reducing overall complexity while maintaining high productivity

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3881234B1Automatically predicting text in images
Publication Date: 2026.05.20 SERVICENOW CANADA INC
  • EP3881234B1 patent drawingFigure 1~2
  • EP3881234B1 patent drawingFigure 3~4
  • EP3881234B1 patent drawingFigure 5~6

AI summary

Systems and methods for detecting and predicting text within images. An image is passed to a feature-extraction module. Each image typically contains at least one text object, and each text object contains at least one character. Based on the image, the feature-extraction module generates at least one feature map indicating text object(s) in the image. The feature map(s) is then passed to a decoder module. In some implementations, the decoder module applies a weighted mask to the feature map(s). Based on the feature map(s), the decoder module predicts a sequence of characters in the text object(s). In some embodiments, that prediction is based on previous known data. The decoder module is directed by a query that indicates at least one desired characteristic of the text object(s). An output module then refines the predicted content. At least one neural network may be used.