Text Field Quality Assessment for Projective OCR Distortion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems struggle with accurately assessing the reliability of text recognition in images with projectively distorted text fields, which can lead to incorrect recognition and potential adverse consequences in applications like identity verification and autonomous vehicles.
Innovation Solution
A method is disclosed that calculates a projective transformation from a restored rectangle to a source quadrangle, constructs an approximation of a minimal scaling coefficient level, and determines if the scaling coefficient satisfies a threshold to assess image quality for reliable text recognition, preventing or reducing the use of low-quality images in the recognition process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Shape
If image restoration is performed on projectively distorted text fields, then the text field geometry is corrected, but the image quality may be insufficient for reliable recognition
Solution Approach 1:
The patent performs preliminary image quality assessment based on projective transformation parameters before conducting text recognition. By calculating the minimal scaling coefficient from the projective transformation matrix H in advance, the system determines whether the restored image quality is sufficient, preventing unreliable recognition attempts on distorted images
2Productivity
If text recognition is performed on all restored images, then processing throughput is maintained, but incorrect recognition increases due to low quality images
Solution Approach 1:
The system performs preliminary quality assessment using the minimal scaling coefficient calculation before text recognition. Images with insufficient quality (scaling coefficient below threshold) are rejected in advance, ensuring that only high-quality images proceed to recognition, thereby maintaining both throughput and accuracy
Solution Approach 2:
The patent introduces an intermediary quality assessment mechanism between image restoration and text recognition. The minimal scaling coefficient serves as a mediator that evaluates whether the restored image quality is sufficient, acting as a gatekeeper that filters out low-quality images before they reach the recognition stage
3Productivity
If the minimal scaling coefficient threshold is lowered to accept more images, then processing volume increases, but recognition accuracy decreases
Solution Approach 1:
The patent establishes a specific threshold range for the minimal scaling coefficient (0.05-0.15) based on empirical analysis. This parameter setting optimizes the balance between processing volume and recognition accuracy by filtering out images with excessive distortion while accepting those with acceptable quality
Data Source
AI summary
Image quality assessment for text recognition in images with projectively distorted text fields. A projective transformation is calculated from a restored rectangle, representing a restored text field, to a source quadrangle, representing a projectively distorted text field in a source image. An approximation of a curve of a minimal scaling coefficient level on a plane corresponding to the restored rectangle is constructed, based on calculations of a discriminant of the curve. When the approximation intersects a representation of the restored rectangle, a restoration of the source image is determined to have insufficient image quality for reliable text recognition. When the approximation does not intersect the representation of the restored rectangle, a minimal scaling coefficient is calculated at a point inside the restored rectangle, and a determination of whether or not the restoration of the source image has sufficient image quality is made based on the minimal scaling coefficient.