Multi-Orientation Text Recognition in Electronic Documents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional OCR systems can only recognize text characters in a predominant orientation, failing to accurately identify text characters in regions with different orientations within the same image, which limits their effectiveness in processing documents with text at multiple orientations.
Innovation Solution
A method that identifies and recognizes text characters at multiple orientations within an image by analyzing regions with different orientations, generating text objects for each orientation, and modifying the electronic document to include these objects, allowing for text search and rendering across various orientations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional OCR process identifies predominant orientation and recognizes text characters having the predominant orientation, then the OCR process can efficiently process text in a single orientation, but it fails to accurately identify text characters in regions with different orientations
Solution Approach 1:
The image is divided into multiple regions, each potentially containing text at different orientations. The system identifies region boundaries and processes each region independently, allowing accurate recognition of text characters in each orientation-specific region while maintaining overall system efficiency.
Solution Approach 2:
The system adds an orientation dimension to the traditional OCR approach by determining the orientation of text in each region and rotating or transforming the image/region accordingly. This allows the OCR engine to handle multiple orientations by transforming them into a common reference frame.
2Adaptability or versatility
If the OCR system processes text at multiple orientations by analyzing different regions, then it can recognize text characters in various orientations, but the processing complexity and computational resources increase
Solution Approach 1:
By segmenting the image into regions with similar text orientations, the system reduces the complexity of processing the entire image at once. Each region can be processed with simpler, orientation-specific algorithms, and the segmentation itself provides structural organization that simplifies subsequent processing steps.
Solution Approach 2:
The system performs preliminary actions by identifying text regions and determining their orientations before performing the actual character recognition. This preliminary analysis includes calculating aspect ratios, generating histograms, and identifying text baselines to prepare the data for more efficient subsequent processing.
3Measurement precision
If the system divides the image into multiple regions and processes each region separately, then it can accurately recognize text at different orientations, but the processing time increases
Solution Approach 1:
The image is segmented into regions based on orientation characteristics, allowing parallel processing of multiple regions. By identifying region boundaries upfront and processing each region independently, the system can optimize processing for each region and potentially process regions in parallel, reducing overall processing time while maintaining accurate orientation detection.
Solution Approach 2:
The system applies partial processing strategies by first analyzing sample regions to determine predominant orientations, then applying orientation-specific processing only where needed. Not every region requires full multi-orientation analysis, and the system can skip certain detailed analysis steps for regions that are clearly identified as having a single dominant orientation.
Data Source
AI summary
Systems, methods, and apparatus, including software tangibly stored on a computer readable medium, involve identifying text in an electronic document. An electronic document that includes an image object is received. In a first region of the image object, a first set of text characters having a first orientation are recognized. In a second region of the image object, a second set of text characters having a second orientation are recognized. The electronic document is modified to include a first text object identifying the first set of text characters and a second text object identifying the second set of text characters. Each identification of text characters includes a set of values that each represent an individual text character recognized in the corresponding region.