Document Image Analysis Feature Extraction Reuse
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing systems experience inefficiencies due to redundant processes in feature extraction and enhancement, where noise removal and low-level segmentation information are generated twice, leading to unnecessary computational overhead.
Innovation Solution
The proposed solution reuses the cleaner image and low-level segmentation information generated during feature extraction in the content adaptive enhancement and segmentation processes, respectively, to eliminate the need for these components to regenerate them independently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If noise removal process is performed during feature extraction to generate a cleaner image, then image quality is improved, but redundant processing occurs when the enhancement stage also creates a cleaner image
Solution Approach 1:
The patent merges the cleaner image generation function from the enhancement stage with the noise removal process in feature extraction. The cleaner image produced during feature extraction is reused by the enhancement stage, eliminating redundant processing while maintaining image quality improvements.
Solution Approach 2:
The noise removal process performs the cleaner image generation action in advance during feature extraction, before the enhancement stage. This preliminary action allows the enhancement stage to reuse the already-cleaned image without performing its own cleaning operation, thus improving productivity.
2Measurement precision
If neighborhood analysis and cleanup phase creates low-level segmentation information during feature extraction, then feature extraction is enhanced, but redundant processing occurs when segmentation and classification also generates this information
Solution Approach 1:
The patent merges the low-level segmentation information generation function from the segmentation and classification stage with the neighborhood analysis and cleanup phase in feature extraction. The segmentation information produced during feature extraction is reused by the segmentation and classification process, eliminating redundant processing while maintaining feature extraction accuracy.
3Manufacturing precision
If independent cleaner image generation is performed in the enhancement stage, then image quality is ensured, but computational overhead increases
Solution Approach 1:
The cleaner image generation capability is made universal by having the feature extraction stage produce a cleaner image that serves dual purposes: it is used for primitive calculation in feature extraction and reused for content adaptive enhancement. This multi-functionality eliminates the need for separate cleaner image generation in the enhancement stage, reducing computational overhead while ensuring image quality.
Data Source
AI summary
Methods, systems and computer program products to improve the efficiency and computational speed of an image enhancement process. In an embodiment, information that is generated as interim results during feature extraction may be used in a segmentation and classification process and in a content adaptive enhancement process. In particular, a cleaner image that is generated during a noise removal phase of feature extraction may be used in a content adaptive enhancement process. This saves the content adaptive enhancement process from having to generate a cleaner image on its own. In addition, low-level segmentation information that is generated during a neighborhood analysis and cleanup phase of feature extraction may be used in a segmentation and classification process. This saves the segmentation and classification process from having to generate low-level segmentation information on its own.