Text Image Trimming via Straight Line Detection and Text Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for trimming text images from backgrounds, especially using digital cameras, suffer from low accuracy due to distortion and inclusion of irrelevant content, requiring manual intervention and relying on straight line detection which can misidentify text edges.
Innovation Solution
An automated text image trimming method that uses straight line detection to identify candidate quadrangles, verifies the edges of the text region using text recognition, and corrects them to increase accuracy, reducing manual effort and data loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If straight line detection method is used to detect text region edges, then the trimming process is automated, but the detection accuracy decreases because straight lines in text may be misidentified as edges
Solution Approach 1:
The detection process is divided into two independent stages: first, straight line detection identifies candidate quadrangle edges; second, text recognition verifies and corrects these edges. This segmentation allows each stage to specialize - geometric detection for speed and text verification for accuracy - resolving the contradiction between automation and precision.
Solution Approach 2:
The text recognition results are fed back to verify and correct the straight line detection results. The system uses the detected candidate quadrangle as a reference frame, then checks whether actual text boundaries align with this frame, providing feedback that corrects misidentifications and improves detection accuracy while maintaining automation.
2Measurement precision
If manual trimming method is used, then the trimming accuracy is high because user can precisely position the text region, but the operation complexity and time consumption increase
Solution Approach 1:
The system performs automatic detection and verification without requiring user intervention. The computer executes the entire trimming process autonomously by combining straight line detection with text recognition verification, eliminating manual operation while maintaining high accuracy through the dual-stage verification mechanism.
3Productivity
If only straight line detection is used, then the processing speed is fast, but the reliability of text region detection decreases due to misidentification of text lines as edges
Solution Approach 1:
The system performs preliminary straight line detection to quickly identify candidate quadrangle edges before conducting text recognition verification. This preliminary action captures the majority of the processing speed benefit while the subsequent verification step ensures reliability, allowing the system to maintain fast processing without sacrificing detection accuracy.
Solution Approach 2:
The candidate quadrangle detected by straight line detection serves as an intermediary reference frame that guides the text recognition process. This intermediary structure allows the system to leverage both the speed of geometric detection and the accuracy of text recognition, with the candidate quadrangle acting as a bridge between the two methods.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Disclosed is a text image trimming method. The method comprises the following steps: step 110, obtaining text image data; step 120, using straight line detection method to detect the straight lines of the text image, obtaining edges of a trimmed quadrangle; step 130, detecting text on the image data, obtaining the coordinates of the boundary points of a text regioon; and step 140, obtaing the final trimming result according to the results of steps 120 and 130. The method can automatically detect the edges of the text region and utilize the detected text region to verify and removes unrelated redundant information thereby, allowing the user to only see the portion containing the text retion useful to the user when viewing image data.