Image Registration Using Character Position Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image registration methods for text recognition, such as in OCR technology, are complex and lack high accuracy, particularly in simplifying steps and enhancing universality.
Innovation Solution
A method and apparatus that recognize characters in an original image, match them with a template image's layout structured region, acquire a projective transformation matrix, and register the image to improve recognition accuracy without needing corner detection or key region detection, using labeled regions for enhanced matching and recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional corner detection or key region detection is used for image registration, then the registration can be performed, but the steps are complex and accuracy is not high
Solution Approach 1:
The patent extracts and removes the complex corner detection and key region detection steps from the traditional image registration process. Instead of performing these complex operations, the method directly uses character position information from template matching to establish correspondence points, thereby simplifying the registration steps while maintaining or improving accuracy.
Solution Approach 2:
The patent makes the image registration method universal by applying it to OCR scenarios without requiring specific corner detection or key region detection algorithms. The method uses general template matching and character position extraction that can work across different document types and layouts, enhancing the universality of the registration process.
2Measurement precision
If template matching with labeled regions is used, then matching accuracy and universality are enhanced, but it requires a pre-prepared template image with labeled layout structured regions
Solution Approach 1:
The patent applies preliminary action by pre-preparing template images with labeled layout structured regions before the actual OCR process. This preprocessing step creates a reference framework that guides the matching process, enabling accurate registration and recognition. The template includes predefined character positions and layout structures that are established in advance.
3Measurement precision
If character recognition and position extraction is performed on the original image, then accurate character positions are obtained, but it increases processing time before registration
Solution Approach 1:
The patent merges the character recognition and position extraction steps with the template matching process. Instead of performing separate recognition and position extraction operations on the original image before registration, the method combines these functions into the matching process itself, where character positions are extracted directly from the matched template regions, thereby reducing overall processing time while maintaining position accuracy.
Data Source
AI summary
The present disclosure discloses a method and an apparatus for processing image information. A specific implementation of the method comprises: recognizing each character in an original image and acquiring a position of the each character; matching a character in the original image with a character in a layout structured region of a template image, and recording identical characters or character strings in the original image and the template image as a matching point pair; acquiring a projective transformation matrix between the matching point pairs according to the position of the character in the original image and the position of the character in the layout structured region of the template image; registering the original image according to the projective transformation matrix to acquire a registered image; and recognizing the registered image to acquire a recognition result. This implementation simplifies steps of image matching in character recognition, enhances matching accuracy and universality, and reduces cost of development.