Text Extraction Pre-processing Technique Ranking System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pre-processing techniques for text extraction from images are not universally applicable due to differences in image properties, leading to inaccurate extraction and inefficiencies in identifying the optimal technique for each image.
Innovation Solution
A method and system that receive an input image and expected text, apply multiple pre-processing techniques, record processing time, compare extracted text with the expected text, and rank techniques based on matching percentage and time to identify the optimum pre-processing technique for text extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single pre-processing technique is applied to all images, then the processing time is reduced and the system is simpler, but the text extraction accuracy deteriorates because image properties vary across different images
Solution Approach 1:
The system dynamically selects pre-processing techniques based on the specific properties of each input image. Instead of using a static, one-size-fits-all approach, the system adapts the pre-processing pipeline to match the characteristics of individual images, thereby maintaining high extraction accuracy across diverse image types while managing processing time through intelligent selection rather than brute-force trial of all techniques.
Solution Approach 2:
The system changes the parameters of the pre-processing technique selection based on image properties. By analyzing characteristics such as image quality, text density, and visual complexity, the system adjusts which pre-processing techniques are applied and in what order, optimizing the balance between extraction accuracy and processing efficiency for each specific image.
2Manufacturing precision
If multiple pre-processing techniques are tried on each image to find the optimum, then the text extraction accuracy is improved, but the processing time increases
Solution Approach 1:
The system performs preliminary analysis of image properties before applying pre-processing techniques. By pre-assessing characteristics such as image quality, text density, and visual complexity, the system can predict which techniques are likely to be effective and apply only those, avoiding the time-consuming trial of all possible techniques while still achieving high extraction accuracy.
Solution Approach 2:
The system uses feedback from the preliminary image analysis to guide the selection and ordering of pre-processing techniques. The feedback mechanism allows the system to adjust the pre-processing pipeline based on the specific needs of each image, ensuring that the most appropriate techniques are applied first, thereby reducing overall processing time while maintaining accuracy.
3Manufacturing precision
If image properties are analyzed in detail before selecting pre-processing techniques, then the extraction accuracy is improved, but the overall processing time increases due to additional analysis steps
Solution Approach 1:
The system performs only the necessary level of image analysis required to make effective pre-processing technique selections, rather than conducting exhaustive analysis of all possible image properties. By focusing on the most critical characteristics such as image quality, text density, and visual complexity, the system achieves sufficient accuracy for technique selection without the time overhead of complete image characterization.
Data Source
AI summary
Disclosed is a system for identifying at least one optimum pre-processing technique for text extraction. A data receiving module receives an input image and an expected text pertaining to the input image. Furthermore, a performance module performs a plurality of pre-processing techniques on the input image in order to extract textual information, present in the input image, by using at least one text extraction technique. Subsequently, a recording module records processing time spent on each pre-processing technique performed on the input image. Further, a comparison module compares the textual information with the expected text. Further to comparing the textual information with the expected text, the ranking module may rank each of the plurality of pre-processing techniques based on a result of comparison between the textual information and the expected text, thereby identifying at least one optimum pre-processing technique for text extraction.