Object Detection Segmentation for Reverse and Vertical Text OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional character recognition systems struggle with accuracy in identifying text glyphs in digital images, particularly in special cases such as reverse text, vertical text, and atypically-sized text.
Innovation Solution
The system employs a machine learning approach that segments digital images into text and non-text regions using an object detection model. It then utilizes an edge-based adaptive binarization model to identify and modify reverse text, vertical text, and atypically-sized text glyphs, converting them into formats more recognizable by optical character recognition models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional character recognition systems use standard OCR models with fixed processing pipelines, then processing speed is maintained, but accuracy deteriorates for special text cases (reverse text, vertical text, atypically-sized text)
Solution Approach 1:
The system segments the text processing pipeline into distinct modules: object detection module for region identification, adaptive binarization module for threshold processing, and OCR module for character recognition. This segmentation allows each module to be optimized independently, improving overall accuracy without requiring complete system redesign
Solution Approach 2:
The system performs preliminary processing steps before OCR, including object detection to identify text regions, adaptive binarization to enhance contrast, and orientation detection to determine text direction. These preliminary actions prepare the image data in advance, enabling the OCR module to process special text cases more accurately
2Adaptability or versatility
If conventional systems extract glyphs in left-to-right fashion according to ordinary reading arrangement, then processing simplicity is maintained, but adaptability deteriorates for non-conventional text arrangements
Solution Approach 1:
The system dynamically adjusts the text extraction and processing approach based on detected text orientation and arrangement characteristics. Instead of fixed left-to-right processing, the system adapts its processing direction and methodology according to the actual text layout, enabling handling of vertical, horizontal, and rotated text
Solution Approach 2:
The system changes processing parameters such as reading direction, extraction orientation, and binarization thresholds based on detected text characteristics. This parameter adaptation allows the same system to handle diverse text arrangements without requiring separate processing pipelines for each case
3Reliability
If conventional systems are designed to identify dark text against light background, then recognition reliability is maintained for standard cases, but reliability deteriorates for reverse text regions
Solution Approach 1:
The system performs preliminary adaptive binarization processing to enhance contrast and distinguish text from background before OCR processing. This preliminary action equalizes the appearance of different text types (standard and reverse text), enabling the OCR module to process both types with the same reliability
Solution Approach 2:
The adaptive binarization module acts as an intermediary between the input image and the OCR module. It transforms the image data into an optimized format that is suitable for OCR processing, mediating the differences between standard and reverse text cases and presenting a unified input to the OCR system
4Measurement precision
If conventional systems use fixed size ranges for text glyphs, then processing efficiency is maintained, but accuracy deteriorates for text outside the size range
Solution Approach 1:
The system dynamically adjusts the size threshold parameters based on the detected text region characteristics and image resolution. Instead of using fixed size ranges, the system adapts its size criteria to the specific context, enabling accurate recognition of text glyphs across varying sizes while maintaining processing efficiency through context-aware parameter adjustment
Data Source
AI summary
The present disclosure relates to systems, methods, and non-transitory computer readable media for accurately enhancing optical character recognition with a machine learning approach for determining words from reverse text, vertical text, and atypically-sized text. For example, the disclosed systems segment a digital image into text regions and non-text regions utilizing an object detection machine learning model. Within the text regions, the disclosed systems can determine reverse text glyphs, vertical text glyphs, and/or atypically-sized text glyphs utilizing an edge based adaptive binarization model. Additionally, the disclosed systems can utilize respective modification techniques to manipulate reverse text glyphs, vertical text glyphs, and/or atypically-sized glyphs for analysis by an optical character recognition model. The disclosed systems can further utilize an optical character recognition model to determine words from the modified versions of the reverse text glyphs, the vertical text glyphs, and/or the atypically-sized text glyphs.