OCR Accuracy in Overlapping Image Series via Coordinate Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optical character recognition (OCR) methods face challenges in accurately processing images with defects such as visual noise, poor focus, and partial overlap, which can lead to suboptimal character recognition and inability to handle documents that cannot be fit into a single image without significant loss of quality.
Innovation Solution
The method involves receiving a series of overlapping images, performing optical symbol recognition, identifying textual artifacts with low frequency occurrence, constructing coordinate transformations, and associating OCR text across images to reconstruct the original document layout, thereby improving OCR quality and handling partial document images effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single image is used for OCR, then the processing speed is fast, but the OCR accuracy deteriorates when the document cannot be fit into a single image without significant loss of quality
Solution Approach 1:
The patent divides the original document into multiple overlapping image fragments when a single image cannot capture the entire document without quality loss. Each fragment is processed separately through OCR, and the results are then combined using coordinate transformation and clustering algorithms to reconstruct the complete text with accurate positioning.
2Measurement precision
If multiple overlapping images are processed, then the OCR accuracy is improved, but the processing time increases
Solution Approach 1:
The patent performs preliminary coordinate transformation and clustering operations to organize multiple OCR results before final text reconstruction. By pre-processing the spatial relationships between image fragments and grouping corresponding text segments, the system efficiently merges results from multiple images without requiring exhaustive processing of all possible combinations.
3Measurement precision
If images with visual noise and poor focus are processed using traditional OCR, then the processing is simple, but the recognition accuracy deteriorates
Solution Approach 1:
The patent combines multiple OCR results from overlapping image fragments, using coordinate transformation and clustering to merge corresponding text segments. This consolidation process allows the system to select the most accurate recognition results from different images, thereby improving overall character recognition accuracy even when individual images contain visual noise or poor focus.
Data Source
AI summary
Systems and methods are described for receiving a current image that partially overlaps with a previous image of a series of images of an original document; performing optical character recognition (OCR) of the current image, producing an OCR text and a corresponding text layout; identifying textual artifacts in the current and previous images, each represented by a sequence of symbols having a frequency of occurrence within the OCR text below a threshold frequency; identifying corresponding base points associated with textual artifacts; identifying parameters of a coordinate transformation converting coordinates of the previous image into coordinates of the current image; associating part of the OCR text with a cluster of symbol sequences, wherein the symbol sequences are produced by processing previously received images; identifying an order of clusters of symbol sequences reflecting a layout of the original document; and producing a resulting OCR text representing a portion of the original document.