Text Localization in Images via Line and Graphic Elimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for text localization in images, particularly those with complex backgrounds or graphics, face challenges in accurately identifying and extracting text due to issues like superimposed text on natural images or complex graphics, leading to incomplete optical character recognition (OCR).
Innovation Solution
The approach involves a processor-implemented method that removes lines and graphics from images using geometric manipulations and the Hough transform, followed by horizontal and vertical projections to identify and extract text lines based on run-lengths of consecutive 1's and 0's in flag vectors, enabling effective text localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing OCR methods are used on images with complex backgrounds and graphics, then processing speed is maintained, but text localization accuracy deteriorates due to inability to separate text from background elements
Solution Approach 1:
The patent segments the image processing task into distinct stages: first removing lines using geometric manipulations, then removing graphics using Hough transform, and finally performing OCR on the cleaned image. This segmentation allows each stage to focus on specific elements, improving text localization accuracy by isolating text from interfering background elements.
Solution Approach 2:
The patent performs preliminary actions by removing lines and graphics before executing the main OCR task. By pre-processing the image to eliminate background elements that would interfere with text recognition, the system improves subsequent OCR accuracy without requiring complex integration during the recognition phase itself.
2Reliability
If text is superimposed on complex backgrounds with natural images or graphics, then image information completeness is maintained, but OCR recognition rate deteriorates
Solution Approach 1:
The patent extracts and removes specific interfering elements (lines and graphics) from the image while preserving the text and background information needed for OCR. By selectively removing only the elements that interfere with text recognition using geometric manipulations and Hough transform, the system improves OCR reliability without unnecessarily losing background context.
3Measurement precision
If traditional text localization methods are used on images with lines and graphics, then processing simplicity is maintained, but text extraction accuracy deteriorates due to interference from background elements
Solution Approach 1:
The patent divides the complex image processing task into simpler sequential sub-tasks: line removal, graphic removal, and text extraction. Each sub-task uses specialized techniques optimized for that specific element type, improving overall text extraction accuracy while keeping each individual processing step relatively simple and manageable.
Solution Approach 2:
The patent performs preliminary removal of lines and graphics before the main text extraction operation. This pre-processing simplifies the subsequent text extraction task by eliminating interfering elements, making the overall process more effective without requiring excessively complex algorithms in the final extraction stage.
Data Source
AI summary
System and method for text localization in images are disclosed. In an embodiment, a line and graphic eliminated image is received. Further, horizontal projection is performed on rows of the image to obtain a first flag vector, the flag vector indicates whether there is text in each row. Furthermore, a number of run-lengths of consecutive 1's and 0's is computed in the first flag vector. Moreover, text lines is extracted in the image based on the computed number of run-lengths of consecutive 1's and 0's in the first flag vector. Also, vertical projection is performed on the text lines to obtain a second flag vector for the text lines. Further, a number of run-lengths of consecutive 1's and 0's is computed in the second flag vectors. Furthermore, text is localized in the image based on the computed number of run-lengths of consecutive 1's and 0's in the second flag vectors.