OCR Region Boundary Validation for Mobile Text Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Electronic devices often erroneously detect non-character regions as character recognition regions, leading to inaccurate character recognition due to reliance on dominant edges, which can be exacerbated by objects within documents.
Innovation Solution
The electronic device employs edge and line detection techniques to identify character recognition regions, compares region sizes with predetermined thresholds, and adjusts virtual reference objects based on image characteristics to validate the region's validity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the character recognition region is detected based on the most dominant edge, then the detection process is simple and fast, but the character recognition region may not be properly detected and objects inside the document may be erroneously detected as the character recognition region
Solution Approach 1:
The patent segments the character recognition process into multiple stages: first detecting candidate regions using edge detection, then validating each candidate by checking for character presence using template matching or neural networks. This segmentation allows simple initial detection followed by precise verification, resolving the contradiction between speed and accuracy.
Solution Approach 2:
The patent introduces an intermediary validation step between region detection and character recognition. This intermediary process checks whether detected regions actually contain characters by comparing against character templates or using neural network classification, preventing erroneous detection of non-character objects while maintaining efficient initial detection.
2Measurement precision
If edge detection and line detection techniques are used to detect the boundary of the character recognition region, then the detection accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent applies preliminary edge detection and line detection to quickly identify potential character recognition regions before performing more complex validation. By preparing candidate regions in advance using these detection techniques, the system improves boundary accuracy while managing complexity through staged processing.
Solution Approach 2:
The patent applies different processing qualities to different regions: simple edge and line detection is applied to the entire image to find candidates, while more complex character validation is applied only to the detected candidate regions. This local quality approach improves overall accuracy without uniformly increasing complexity across the entire processing pipeline.
3Loss of information
If the entire image is selected for character recognition when the detected region size is less than the predetermined size, then no character information is lost, but the recognition processing time increases
Solution Approach 1:
The patent uses a size threshold as a decision criterion: when the detected region is smaller than the threshold, it selectively processes only that region (partial action), but when the threshold is not met, it processes the entire image (excessive action) to ensure no character information is lost. This approach balances information completeness with processing efficiency based on detected region characteristics.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Various embodiments of the present disclosure relate to an electronic device and a character recognition method thereof. The electronic device may include a housing, a touchscreen display exposed through a first portion of the housing, an image sensor exposed through a second portion of the housing, a wireless communication circuit located in the housing, a processor operatively coupled to the touchscreen display, the image sensor, and the wireless communication circuit, and a memory operatively coupled to the processor. The memory may store instructions, which when executed, cause the processor to control the electronic device to display a user interface including an image acquired from the image sensor or stored in the memory, and recognize a region including a text in the image for optical character recognition by detecting a boundary of the region, comparing a size of the region with a predetermined size, and selecting the region for the optical character recognition based at least in part on the comparison result. Other various embodiments are also possible.