Vector-Encoded Text Region Selection for Faster OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical character recognition (OCR) processes for digital images are inefficient and computationally expensive due to the need to analyze the entire document, leading to increased latency and resource usage.
Innovation Solution
A vector encoder is trained to encode images representing text into vectors, grouping similar text closely in a vector space and dissimilar text far apart, allowing for the identification of specific regions within an image that match a search query, thereby applying OCR only to relevant sections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard OCR process is applied to the entire document image, then complete text recognition is achieved, but processing time and computational resources increase significantly
Solution Approach 1:
The patent divides the document image into multiple sub-images or regions of interest before applying OCR. By segmenting the full document into smaller manageable parts, the system can process only relevant sections, significantly reducing processing time while maintaining recognition accuracy for the segmented regions
Solution Approach 2:
The patent extracts and identifies specific regions containing text of interest from the full document image using preliminary analysis techniques. By taking out only the relevant text regions for OCR processing, the system avoids the computational overhead of processing the entire document, thus reducing latency while preserving recognition quality
2Productivity
If standard OCR process is applied to the entire document image, then all text is converted to searchable format, but computational resources and processing cost increase
Solution Approach 1:
The patent segments the document into multiple sub-images and applies OCR selectively to these segments rather than the entire document. This segmentation approach increases productivity by enabling parallel processing of multiple regions while reducing computational resource consumption by excluding irrelevant areas from OCR processing
Solution Approach 2:
The patent applies partial action by performing OCR only on identified text regions rather than the complete document. This partial processing approach maintains sufficient productivity for the extracted text while significantly reducing the computational energy required compared to full-document OCR
3Speed
If vector encoder is used to group similar text in vector space, then relevant text regions can be identified faster, but initial training and vector computation overhead is introduced
Solution Approach 1:
The patent implements preliminary action by pre-training the vector encoder model and pre-computing vector representations of text regions before actual OCR processing. This preliminary preparation enables rapid identification of relevant text regions during runtime, significantly improving speed while the complexity is managed as a one-time setup cost rather than ongoing operational overhead
Data Source
AI summary
Provided are a computer program product, system, and method for training and using a vector encoder to determine vectors for sub-images of text in an image to subject to optical character recognition. A vector encoder is trained to encode images representing text into vectors in a vector space. Vectors of images representing similar text have a high degree of cohesion in the vector space. Vectors of images representing dissimilar text have a low degree of cohesion in the vector space. An input image is processed to determine sub-images of the input image that bound text represented in the input image. The sub-images are inputted to the vector encoder to output sub-image vectors. The vector encoder generates a search vector for search text. Optical character recognition is applied to at least one region of the input image including the sub-images having sub-image vectors matching the search vector.