Image Processing Apparatus Character Recognition Preprocessing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for improving character recognition accuracy in document scanning, such as those described in Japanese Patent Laid-Open No. H7-6203, often involve unnecessary preprocessing steps that increase processing time and cost without significantly enhancing accuracy.
Innovation Solution
The proposed solution involves an image processing apparatus that performs optimized preprocessing for character recognition by determining whether to apply additional preprocessing based on pre-registered information for similar business form images, thereby reducing unnecessary processing steps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If predetermined preprocessing is performed each time for character recognition, then character recognition accuracy is improved, but processing time and cost increase due to unnecessary processes
Solution Approach 1:
The system performs preprocessing in advance by creating multiple candidate images through various preprocessing methods (rotation, scaling, filtering) and stores them as preprocessed candidates. When a character recognition request comes, the system selects from preprocessed candidates rather than performing all preprocessing operations in real-time, thus improving speed while maintaining accuracy options.
Solution Approach 2:
The system performs only the necessary preprocessing operations based on the specific input characteristics. By analyzing the input image properties first, the system applies only the required preprocessing methods (e.g., only rotation if needed, or only filtering) rather than applying all possible preprocessing operations, reducing unnecessary processing time while maintaining sufficient accuracy.
2Measurement precision
If predetermined preprocessing is performed each time for character recognition, then character recognition accuracy is improved, but processing cost increases due to unnecessary processes
Solution Approach 1:
The system pre-processes images in advance and stores candidate images in a database. When character recognition is needed, the system retrieves preprocessed candidates instead of performing expensive preprocessing operations in real-time, significantly reducing processing cost while maintaining the option to achieve high accuracy through selective preprocessing.
Solution Approach 2:
The system applies only the necessary preprocessing operations based on input characteristics analysis. By avoiding unnecessary preprocessing steps (such as applying rotation to already upright images or complex filtering to clear images), the system reduces computational cost while maintaining sufficient recognition accuracy for the given input.
3Measurement precision
If multiple preprocessing methods are applied to improve accuracy, then character recognition accuracy improves, but device complexity increases
Solution Approach 1:
The system divides the preprocessing into separate, independent modules (rotation module, scaling module, filtering module, etc.). Each module handles a specific preprocessing operation independently. This modular approach allows the system to apply only the necessary preprocessing steps for each input image, reducing overall complexity while maintaining the capability to achieve high accuracy through selective application of multiple methods.
Solution Approach 2:
The system dynamically selects which preprocessing methods to apply based on the characteristics of each input image. Rather than applying a fixed sequence of all preprocessing methods to every image, the system analyzes the input and adaptively chooses the appropriate preprocessing operations, simplifying the processing pipeline for each case while maintaining accuracy when needed.
Data Source
AI summary
The image processing apparatus having a scan function and including: a memory storing a program; and a processor executing the program to set a property of a business form file by using results of character recognition processing for a business form image obtained by reading a business form; perform predetermined preprocessing before the character recognition processing for the business form image; and perform the character recognition processing for each character area within the business form image for which the preprocessing is performed, wherein as the preprocessing, first preprocessing performed as a default and second preprocessing performed additionally for improving accuracy of the character recognition processing exist and in the performing preprocessing, whether to perform the second preprocessing following the first preprocessing is determined based on information specifying contents of preprocessing to be performed, which is registered in advance for a past business form image similar to the business form image.