Image Text Recognition Correction via Segmentation and Language Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in achieving high accuracy for text recognition, especially in complex recognition scenarios, due to variations in text formats and noise.
Innovation Solution
A method involving text recognition, segmentation processing, and correction using a pre-trained language model to improve the accuracy of text recognition results. This method performs text recognition on a target image, segments the recognized text, and corrects it based on text segments obtained from the segmentation processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If text recognition is performed directly on the target image without segmentation, then the processing speed is faster, but the accuracy of text recognition deteriorates in complex scenarios
Solution Approach 1:
The patent applies segmentation by dividing the recognized text into multiple text segments before processing them through the language model. This segmentation approach allows the system to handle complex text structures more effectively, improving recognition accuracy while managing processing complexity through structured breakdown of the text data.
Solution Approach 2:
The patent performs preliminary text recognition and segmentation before applying the language model for correction. This preliminary processing prepares the data in an optimized format, allowing the language model to focus on correction tasks and improving overall accuracy without unnecessarily increasing processing complexity during the correction phase.
2Measurement precision
If a pre-trained language model is used to correct recognized text, then the text recognition accuracy is improved, but the processing time increases
Solution Approach 1:
The patent applies partial action by using the language model specifically for correction purposes rather than performing complete text recognition from scratch. The language model corrects only the recognized text segments, which is less computationally intensive than full recognition, thereby reducing processing time while maintaining accuracy improvements.
Solution Approach 2:
By segmenting the text before correction, the system processes smaller text units through the language model, which reduces the computational burden and processing time compared to correcting entire text blocks, while still achieving accuracy improvements through targeted correction of segmented portions.
3Measurement precision
If text segmentation is performed before correction, then the correction accuracy is improved through better context understanding, but the processing complexity increases
Solution Approach 1:
The patent segments the recognized text into multiple text segments that preserve contextual relationships. This segmentation structure enables the language model to understand context better within each segment while maintaining overall text coherence, improving correction accuracy without requiring overly complex processing architecture.
Solution Approach 2:
The segmentation is performed as a preliminary step before correction, organizing the text data into manageable segments with preserved context. This preliminary organization simplifies the subsequent correction process by providing structured input to the language model, reducing the complexity of the correction operation itself.
Data Source
AI summary
The disclosure relates to a method, apparatus, readable storage medium and electronic device of image processing. The method includes: performing text recognition on a target image, to obtain a recognized text; performing segmentation processing on the recognized text; and obtaining, based on a text segment obtained from the segmentation processing, a target text by correcting the recognized text through a pre-trained language model.