Text Recognition in Images Using Superresolution and Adaptive Preprocessing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing technologies face challenges in accurately recognizing text within images due to low image quality and environmental factors such as varying text size, orientation, and obstructions, which complicates the identification of address locations, business names, and other information in urban scenes.

Innovation Solution

A system and method for text recognition that involves preprocessing images to enhance candidate text regions, aligning and compositing them to create a superresolution image, and performing optical character recognition to improve text identification accuracy, incorporating features like adaptive gray value normalization and three-dimensional range data to distinguish text from background.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional OCR is applied directly to low-quality images, then processing speed is maintained, but text recognition accuracy deteriorates due to image distortions, compression artifacts, and environmental factors

Engineering Contradiction:
Improvetext recognition accuracyVSAvoidimage processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The image is divided into multiple candidate text regions using detection algorithms that identify potential text locations based on features like edges, contours, and local binary patterns. This segmentation isolates text areas from the rest of the image, allowing focused enhancement and recognition processing on specific regions rather than the entire image.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Candidate text regions undergo preprocessing operations including contrast enhancement, noise reduction, and geometric correction before OCR recognition. This preliminary action improves the quality of text regions in advance, making them more suitable for accurate character recognition while maintaining a balance between processing complexity and recognition accuracy.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If multiple image processing operations are performed to enhance text regions, then text recognition accuracy is improved, but processing time increases

Engineering Contradiction:
Improvetext recognition accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

Different enhancement operations are applied selectively to different candidate text regions based on their specific characteristics. Regions with low contrast receive contrast enhancement, regions with noise receive denoising filters, and regions with perspective distortion receive geometric correction. This localized approach improves text recognition accuracy for each region while avoiding unnecessary processing operations.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system applies a set of preprocessing operations to candidate text regions, including some operations that may be redundant for certain regions. This ensures that all potential text regions receive sufficient enhancement to achieve high recognition accuracy, even if some processing steps are more intensive than strictly necessary for each individual region.

Inventive Principle:
Principle #16Partial or excessive action

3Reliability

If text detection sensitivity is increased to capture all potential text, then text identification completeness is improved, but false positive rate increases due to non-text regions being misidentified

Engineering Contradiction:
Improvetext identification completenessVSAvoidfalse positive rate
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The system uses feedback mechanisms where detection results are evaluated and used to adjust detection parameters. Candidate regions that are likely to be false positives are identified through validation checks comparing detected regions against expected text characteristics such as aspect ratios, pixel density patterns, and structural features. This feedback loop refines the detection process to reduce false positives while maintaining high text identification completeness.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

Detection parameters such as sensitivity thresholds, region size constraints, and feature weightings are adjusted based on the specific image characteristics and detection results. By dynamically changing these parameters, the system optimizes the balance between capturing all text regions and minimizing false positive identifications of non-text areas.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2041678B1Recognizing text in images
Publication Date: 2018.10.17 GOOGLE LLC
  • EP2041678B1 patent drawingFigure 1
  • EP2041678B1 patent drawingFigure 2
  • EP2041678B1 patent drawingFigure 3

AI summary

Methods, systems, and apparatus including computer program products for using extracted image text are provided. In one implementation, a computer-implemented method is provided. The method includes receiving an input of one or more image search terms and identifying keywords from the received one or more image search terms. The method also includes searching a collection of keywords including keywords extracted from image text, retrieving an image associated with extracted image text corresponding to one or more of the image search terms, and presenting the image.