Intrinsic Image Segmentation for Document Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems face challenges in accurately performing object recognition and optical character recognition due to varying illumination conditions and distortions in images captured by cellular devices, such as shadows and folds in paper documents.
Innovation Solution
A method and system that analyze digital images for spatio-spectral information to segregate intrinsic material and illumination aspects, allowing for improved image processing tasks like object recognition and optical character recognition by using spatio-spectral operators and constraints to separate and optimize material and illumination components within the images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image processing is performed on images captured under varying illumination conditions, then object recognition and optical character recognition tasks can be executed, but accuracy deteriorates due to shadows and lighting variations
Solution Approach 1:
The patent segments the image into multiple intrinsic components (reflectance map, illumination map, shadow map) to separate the object information from illumination artifacts. This segmentation allows independent processing of each component, enabling accurate object recognition even under varying lighting conditions by removing shadows and illumination variations from the analysis.
Solution Approach 2:
The patent introduces intrinsic images as intermediary representations that serve as mediators between the raw image and the final recognition task. These intrinsic images (reflectance map, illumination map) act as intermediate stages that isolate and process different aspects of the image separately, allowing the system to eliminate harmful illumination effects before performing object or character recognition.
2Ease of operation
If paper documents are scanned using portable devices, then convenience is improved, but measurement accuracy deteriorates due to folds, creases, and distortions
Solution Approach 1:
The patent segments the document image into intrinsic components including a reflectance map that represents the true document appearance and an illumination map that captures folds and creases. By separating these components, the system can process the reflectance map for accurate character recognition while independently analyzing the illumination map to identify and compensate for physical document distortions.
Solution Approach 2:
The patent replaces physical scanning mechanisms with computational image processing. Instead of using rigid, precision mechanical scanners that eliminate document distortions physically, the system uses digital image analysis and intrinsic image processing to computationally remove the effects of folds and creases, achieving similar accuracy with portable devices.
3Speed
If traditional image processing methods are used without intrinsic image segregation, then processing speed is maintained, but reliability deteriorates due to interference from illumination variations
Solution Approach 1:
The patent implements segmentation into intrinsic images (reflectance map, illumination map, shadow map) that can be processed in parallel or sequentially with optimized algorithms. This segmentation allows the system to focus computational resources on the most relevant components for each specific task, maintaining efficiency while improving reliability through targeted processing of illumination-invariant features.
Data Source
AI summary
In a first exemplary embodiment of the present invention, an automated, computerized method is provided for processing an image. According to a feature of the present invention, the method comprises the steps of, at a computer system, receiving an image recorded at and transmitted from a remote source, at the computer system, processing the transmitted image, for segregation of the transmitted image into a corresponding intrinsic image and, at the computer system, processing the corresponding intrinsic image to perform a preselected task, for a result. As a further feature of the first exemplary embodiment of the present invention, the remote source comprises a cell phone or remote computer system.