Dynamic Text Block Detection for Shifted Index Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting indices from scanned images, such as those using OCR processing, face challenges when the position of text blocks shifts due to changes in written content, leading to failed index extraction.
Innovation Solution
An image processing apparatus that detects text blocks in an input image, determines the corresponding registered document, and uses a partial layout defined in the registered document to locate the text block corresponding to a processing target item, even if its position has shifted.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If partial OCR processing is performed on a registered index region, then processing load is reduced, but index extraction fails when text block position shifts
Solution Approach 1:
The patent applies dynamics by making the text block positioning adaptive rather than fixed. The system dynamically adjusts the search region based on detected text block positions, allowing the extraction region to move and resize according to actual document variations. This resolves the contradiction by enabling fast partial OCR processing while maintaining reliability through dynamic position adaptation.
Solution Approach 2:
The patent changes parameters by adjusting the search region coordinates and dimensions based on detected text block positions. Instead of using fixed registered coordinates, the system modifies the search region parameters (position, size) to accommodate shifts in text block locations, thereby maintaining extraction accuracy while still using efficient partial OCR processing.
2Reliability
If OCR processing is performed on the entire scanned image, then index extraction reliability is ensured, but processing load increases
Solution Approach 1:
The patent applies segmentation by dividing the OCR processing into two stages: first detecting text block positions, then performing OCR only on relevant regions. This segmentation approach ensures reliability by accurately identifying text blocks while improving productivity by limiting OCR processing to specific regions rather than the entire image.
Solution Approach 2:
The patent uses partial action by performing OCR processing only on necessary regions (index regions and surrounding text blocks) rather than the entire image. This approach maintains extraction reliability for target indices while reducing overall processing load by excluding unnecessary areas from intensive OCR processing.
3Device complexity
If text block position is fixed in registered document, then processing simplicity is maintained, but extraction fails when position shifts occur
Solution Approach 1:
The patent transforms the fixed position system into a dynamic one by detecting actual text block positions and adjusting search regions accordingly. This dynamic approach maintains reasonable system complexity while significantly improving extraction reliability when text block positions shift due to document variations.
Data Source
AI summary
An image processing apparatus includes at least one memory that stores instructions; and at least one processor that execute the instructions to perform: detecting text blocks in an input image; determining a registered document corresponding to the input image among a plurality of registered documents; determining the text block in the input image that corresponds to a processing target item, based on a partial layout defined in the determined registered document and including a first text block corresponding to the processing target item and at least one second text block present near the first text block; and obtaining a character string corresponding to the processing target item by performing character recognition processing on the determined text block.