Single-Image Multiple Document Extraction with Superpixel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods fail to effectively handle multiple documents copied, photographed, or scanned onto a single image frame, leading to manual processing and resource consumption.
Innovation Solution
A neural network-based system that generates a superpixel segmentation map, identifies connected components, and creates image dividing lines to separate multiple documents within a single image, utilizing a downscale, context, and final classification blocks with rectifier activation functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional methods are used to process a single image frame, then the processing is simple and fast, but multiple documents cannot be effectively separated and require manual processing
Solution Approach 1:
The patent applies segmentation by dividing the single image frame into multiple document regions using superpixel segmentation maps. The system segments the image by identifying connected components and generating binary masks that separate individual documents, enabling automatic extraction of multiple documents from a single frame without manual intervention.
2Productivity
If manual processing is used for multiple documents on a single frame, then processing accuracy is maintained, but resource consumption increases and efficiency decreases
Solution Approach 1:
The system implements self-service by using trained neural networks to automatically perform document separation and identification tasks that would otherwise require manual processing. The neural network autonomously analyzes the image frame, generates segmentation maps, identifies connected components, and extracts individual documents without human intervention, thereby improving productivity while reducing resource consumption.
3Measurement precision
If a neural network with multiple blocks is used for document separation, then automatic extraction of multiple documents is achieved, but computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by using a downscale block to reduce the image resolution before processing. This preliminary downsampling reduces the computational complexity of subsequent processing steps while maintaining sufficient accuracy for document separation. The system performs the segmentation and classification operations on the downsampled image, then maps the results back to the original image space, thereby reducing processing time while maintaining measurement precision.
Data Source
AI summary
System and method for document image detection, comprising: producing, using a neural network, a superpixel segmentation map of an input image; generating a superpixel binary mask by associating each superpixel of the superpixel segmentation map with a class of a predetermined set of classes; identifying one or more connected components in the superpixel binary mask; for each connected component of the superpixel binary mask, identifying a corresponding minimum bounding polygon; creating one or more image dividing lines based on the minimum bounding polygons; and defining boundaries of one or more objects of interest based on at least a subset of the image dividing lines.