Image Character Localization Through Connected Component Association

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing character recognition techniques, particularly those using neural networks, struggle to accurately estimate the position of recognized characters in images, especially in low-quality or perspective-distorted scenarios, leading to unreliable character localization.

Innovation Solution

An image processing system and method that utilizes a neural network for character recognition, followed by associating recognized characters with connected components in the image, allowing for precise localization by enhancing the position estimation through user corrections and augmentation, thereby improving the accuracy of character positioning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If neural network is used for character recognition, then character recognition accuracy is improved, but position estimation accuracy deteriorates

Engineering Contradiction:
Improvecharacter recognition accuracyVSAvoidposition estimation accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent segments the character recognition process into two independent stages: first, the neural network performs character recognition to identify what characters are present; second, connected component analysis separately determines the positions of individual character components in the image. This segmentation allows each stage to optimize for its specific function without compromising the other.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces connected components as an intermediary element that bridges character recognition and position estimation. The connected components are extracted from the image and associated with recognized characters, serving as a mediator that provides accurate position information without requiring the neural network to directly output position data.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If character recognition is performed on low quality images with perspective issues, then recognition robustness is improved, but position estimation reliability deteriorates

Engineering Contradiction:
Improverecognition robustnessVSAvoidposition estimation reliability
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent separates the recognition function from the positioning function. The neural network processes the entire image to recognize characters regardless of quality or perspective, while connected component analysis separately identifies positionable elements in the original image, ensuring position estimation remains reliable even when recognition operates on degraded input.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing qualities to different aspects of the problem: the neural network performs robust recognition on the full image with all its imperfections, while the connected component analysis operates on local regions to identify precise positions of individual character elements, maintaining high position reliability despite overall image quality issues.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3776332B1An image processing system and an image processing method
Publication Date: 2025.07.16 CANON EURPOPA NV
  • EP3776332B1 patent drawingFigure 1
  • EP3776332B1 patent drawingFigure 2
  • EP3776332B1 patent drawingFigure 3A

AI summary

An image processing system and an image processing method for localising recognised characters in an image. An estimation means is configured to estimate a first location of a recognised character that has been obtained by performing character recognition of the image. A determination means is configured to determine second locations of a plurality of connected components in the image. A comparison means is configured to compare the first location and the second locations, to identify a connected component associated with the recognised character. An association means is configured to associate the recognised character, the identified connected component, and the second location of the identified connected component.