Curved Text Recognition Using Polygon-to-Rectangle Correction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional text recognition methods for curved text suffer from low detection and recognition accuracy due to the imprecision of shape robust text detection algorithms.

Innovation Solution

A text recognition method that involves detecting a polygonal area around the curved text, correcting it to multiple rectangular areas, and performing text recognition on these areas to improve accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If shape robust text detection with progressive scale expansion network algorithm is used to detect curved text, then text detection can be performed, but detection accuracy is low

Engineering Contradiction:
Improvecurved text detection accuracyVSAvoidrecognition accuracy
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the curved text detection problem into multiple steps: first detecting the curved text region, then fitting a curve to the text centerline, and finally correcting the detection area based on the fitted curve. This segmentation allows each step to be optimized independently, improving overall detection accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary curve fitting on the detected text centerline before performing the final recognition. By pre-processing the detection area through curve fitting and correction, the system prepares the data in advance to reduce distortion and improve recognition accuracy in the subsequent step.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If conventional text detection algorithms are used, then detection process is simple, but detection accuracy is low

Engineering Contradiction:
Improvetext detection accuracyVSAvoiddetection algorithm complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The detection process is segmented into distinct modules: initial detection using a standard algorithm, curve fitting on the detected centerline, and detection area correction based on the fitted curve. This modular segmentation maintains relative simplicity while improving accuracy through multiple processing stages.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a curve fitting dimension to the traditional detection process. By fitting a curve to the text centerline and using this curve to correct the detection area, the system adds a geometric transformation layer that improves accuracy without requiring a completely complex detection algorithm.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Reliability

If curved text is recognized directly from detection results, then recognition process is fast, but recognition accuracy is low due to text distortion

Engineering Contradiction:
Improvecurved text recognition accuracyVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent performs preliminary correction of the detection area using the fitted curve before recognition. By pre-distorting the detection area to compensate for the curved text layout, the system reduces the distortion that would otherwise occur during recognition, thereby improving accuracy without requiring complex real-time correction during the recognition phase.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The fitted curve acts as an intermediary between the detected text region and the final recognition process. The curve is used to generate correction parameters that transform the detection area, serving as a bridge that reduces distortion while maintaining processing efficiency.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4064123B1Text recognition method and apparatus
Publication Date: 2026.03.25 HUAWEI TECH CO LTD
  • EP4064123B1 patent drawingFigure 1a
  • EP4064123B1 patent drawingFigure 1b
  • EP4064123B1 patent drawingFigure 2~3

AI summary

A text recognition method and apparatus that relate to the field of information processing technologies are provided. This effectively resolves a problem of a low recognition rate of curved text. The text recognition method includes: obtaining a to-be-detected image; determining a target text detection area in the to-be-detected image, where the target text detection area includes target text in the to-be-detected image, and the target text detection area is a polygonal area including m (m is a positive integer greater than 2) vertex pairs; correcting the polygonal area to m - 1 rectangular areas to obtain a corrected target text detection area; and performing text recognition on the corrected target text detection area, and outputting the target text.