Text Rectifier Using TILT for Deformed Image OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current OCR technologies are sensitive to character deformations such as rotation and skew, leading to decreased recognition accuracy when dealing with images of text captured from varying viewpoints, especially in cases of individual characters or short phrases where global text layout assumptions are not met.
Innovation Solution
The Text Rectifier employs a Transform Invariant Low-Rank Texture (TILT) extraction process using rank minimization techniques to remove affine and projective transforms from images of text, allowing for the rectification of text at arbitrary camera viewpoints, and applies binarization and inversion to enhance OCR accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional OCR systems are used, then recognition accuracy is high for standard upright text, but recognition accuracy degrades rapidly with rotation or skew
Solution Approach 1:
The system performs preliminary rectification of distorted text images before OCR recognition. By detecting rotation and skew angles and applying corrective transformations, the text is preprocessed into a standard orientation that OCR systems can accurately recognize, thereby maintaining high recognition accuracy across various input conditions.
Solution Approach 2:
The system dynamically adjusts geometric parameters (rotation angle, skew value) of the input text image based on detected distortion. By transforming these parameters to return the text to its canonical orientation, the system enables OCR engines to operate within their optimal accuracy range regardless of initial distortion levels.
2Ease of manufacture
If global text layout assumptions are used for rectification, then rectification works well for large text regions, but rectification fails for individual characters or short phrases
Solution Approach 1:
The system segments the input image into multiple text line candidates and processes each segment independently for rectification. By detecting individual text lines or character groups rather than assuming a global layout, the system can accurately rectify small text regions while maintaining effectiveness for larger regions through hierarchical processing.
Solution Approach 2:
The system applies different rectification strategies to different regions of the image based on local text characteristics. For each detected text line or character group, the system computes local rotation and skew parameters tailored to that specific region, enabling accurate rectification of individual characters or short phrases while preserving the ability to handle larger text regions.
Data Source
AI summary
A “Text Rectifier” provides various techniques for processing selected regions of an image containing text or characters by treating those images as matrices of low-rank textures and using a rank minimization technique that recovers and removes image deformations (e.g., affine and projective transforms as well as general classes of nonlinear transforms) while rectifying the text or characters in the image region. Once distortions have been removed and the text or characters rectified, the resulting text is made available for a variety of uses or further processing such as optical character recognition (OCR). In various embodiments, binarization and/or inversion techniques are applied to the selected image regions during the rank minimization process to both improve text rectification and to present the resulting images of text to an OCR engine in a form that enhances the accuracy of the OCR results.