Barcode Readability Enhancement via Frequency-Guided Computer Vision

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing barcode reading technologies struggle with reading low-quality barcodes, which are often unreadable due to poor printing, damage, or environmental deterioration, compromising identity verification and fraud detection processes.

Innovation Solution

A machine learning-based system that uses an object detection convolutional neural network to detect and enhance low-quality barcodes, trained with synthetic personal information to improve readability and accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional barcode reading technologies are used, then the system is simple and easy to operate, but the barcode readability is poor when the barcode is of low quality

Engineering Contradiction:
Improvebarcode readabilityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

A machine learning-based image enhancement model is introduced as an intermediary component between the barcode capture device and the barcode reader. This mediator processes low-quality barcode images by removing noise, enhancing contrast, and restoring degraded regions, thereby improving barcode readability without requiring changes to the core reading system architecture

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system creates an enhanced copy of the low-quality barcode image through machine learning reconstruction. The model generates a synthetic high-quality version of the degraded barcode by learning from training data, allowing the reader to process this improved copy instead of the original low-quality image

Inventive Principle:
Principle #26Copying

2Reliability

If machine learning-based enhancement is applied to improve barcode quality, then the barcode readability increases, but the processing time and computational resources increase

Engineering Contradiction:
Improvebarcode readabilityVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The machine learning model is pre-trained offline using a comprehensive dataset of barcode images with various degradation types. During deployment, only the trained model inference is required, which is significantly faster than training time. The preliminary training phase captures complex degradation patterns, enabling rapid reconstruction of degraded barcodes during actual use

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system dynamically adjusts processing parameters based on the quality assessment of the input barcode image. For mildly degraded barcodes, lighter enhancement filters are applied; for severely degraded barcodes, more intensive reconstruction algorithms are activated. This adaptive approach optimizes processing time based on the actual degradation level

Inventive Principle:
Principle #35Parameter changes

3Reliability

If synthetic personal information is used for training the model, then the model can enhance low-quality barcodes effectively, but the training data requirements and system complexity increase

Engineering Contradiction:
Improvebarcode enhancement accuracyVSAvoidtraining system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

Instead of collecting and storing actual personal information from real barcodes, the system uses synthetic data that replicates the structural and degradation characteristics of real barcodes. The training data consists of generated barcode images with simulated degradation patterns, preserving the essential features needed for accurate enhancement while avoiding privacy concerns and data handling complexity

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The training process utilizes parameter transformations to generate diverse synthetic barcode images from a limited set of template data. By applying various virtual degradation parameters (blur levels, noise intensity, contrast variations), the system creates comprehensive training datasets without requiring actual personal information, reducing data security burdens

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4535225A1Generative system and method for enhancing readability of barcodes using frequency guided computer vision
Publication Date: 2025.04.09 SOCURE INC
  • EP4535225A1 patent drawingFigure 1
  • EP4535225A1 patent drawingFigure 2
  • EP4535225A1 patent drawingFigure 3

AI summary

Provided is a system and methodology which enhances the readability of barcodes using machine learning techniques wherein a model is trained using exemplary data and the resulting model is used in production to increase the chances of successfully capturing the data in a barcode notwithstanding that the barcode may be of low quality. According to the teachings of the present invention, the system is trained using large quantities of synthetic personal information associated with barcodes. Once the system is placed into production, the trained model is applied as against newly identified barcodes to enhance the readability of the original barcode based on features associated with the trained model.