Scanned Image ROI Extraction via Superpixel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing solutions struggle to accurately and efficiently detect and extract the region of interest (ROI) from high-resolution scanned images, particularly those with similar foreground and background colors, requiring high-end computing resources and intensive training data.
Innovation Solution
A processor-implemented method that partitions scanned images into predefined parts, applies filtering techniques, and uses clustering algorithms to identify and extract the ROI, determining the image type based on color intensity and corner points, without relying on pre-defined templates or background colors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning based methods are used for ROI detection, then detection accuracy is improved, but computing resource requirements and training data needs increase significantly
Solution Approach 1:
The image is divided into multiple superpixels using the SLIC algorithm, transforming a complex global detection problem into simpler local region analysis. This segmentation allows the system to process high-resolution images efficiently by working with smaller, manageable regions rather than the entire image at once, reducing computing resource requirements while maintaining detection accuracy.
Solution Approach 2:
The patent introduces superpixels as an intermediary representation between the original image and the final ROI detection. By first converting the image into superpixel regions and then analyzing these intermediate regions for text content, the system achieves accurate detection without requiring the intensive computational resources of direct deep learning approaches on full-resolution images.
2Measurement precision
If high-resolution scanned images are processed, then ROI detection accuracy is improved, but processing time increases
Solution Approach 1:
The high-resolution image is segmented into superpixels, reducing the effective processing load. Instead of analyzing every pixel in the high-resolution image, the system processes grouped superpixel regions, significantly reducing processing time while preserving the ability to detect ROI accurately in the original high-resolution image.
Solution Approach 2:
The patent creates a superpixel representation (a simplified copy) of the high-resolution image for processing. This superpixel copy contains the essential structural information needed for ROI detection but with reduced complexity, allowing fast processing while the final ROI coordinates are mapped back to the original high-resolution image for accurate extraction.
3Productivity
If conventional image processing solutions are used, then processing speed is maintained, but ability to detect exact document region automatically is lost
Solution Approach 1:
The system performs automatic ROI detection by analyzing superpixel regions for text content and document characteristics without requiring manual intervention or pre-defined templates. The algorithm autonomously identifies the document region by processing superpixels and detecting text patterns, achieving both automation and reasonable processing speed.
Solution Approach 2:
The patent transforms the image processing approach by changing from direct pixel-based processing to superpixel-based processing. This parameter change in the representation format enables automatic detection of document regions while maintaining efficient processing speeds, as superpixels provide a coarser-grained structure that is faster to analyze than individual pixels.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
ROI (Region of Interest) detection is an important step in extracting relevant information from a document image. Such images are very high-resolution images in nature and size of images is in order of megabytes, which makes text detection pipeline very slow. Traditional methods detect and extract ROI from images, but these work only for specific image types. Other approaches include deep learning (DL) based methods for ROI detect which need intensive training and require high end computing infrastructure/resources with graphical processing unit (GPU) capabilities. Systems and methods of the present disclosure perform ROI extraction by partitioning input image into parts based on its visual perception and then classify the image in first or second category. Region of interest is extracted from a resized image based on the classification by applying image processing techniques. Further, the system determines whether the input image is a pre-cropped image or a normal scanned image.