Document Image Segmentation Using AI Masks and Bounding Rectangles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to efficiently segment multiple documents within a single image for quick processing, which is crucial for timely document verification during onboarding processes, as delays can lead to user disinterest.
Innovation Solution
An AI-based method involving deep learning models, thresholding, and image processing to identify and isolate individual documents within an image by creating masks, converting to grayscale, applying Otsu thresholding, and using minimum bounding rectangles for cropping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple documents are processed as a single image, then processing time is reduced, but document segmentation accuracy deteriorates
Solution Approach 1:
The patent applies image segmentation techniques to divide a single image containing multiple documents into separate document regions. The system identifies boundaries between documents and creates individual masks for each document, enabling subsequent processing of each document separately while maintaining the efficiency of batch processing.
Solution Approach 2:
The patent introduces an intermediary segmentation layer that acts as a mediator between the input image and final document processing. This intermediary step generates mask images that isolate individual documents, allowing the system to benefit from both efficient batch processing and accurate individual document handling.
2Device complexity
If traditional image processing methods are used, then system complexity is reduced, but segmentation performance deteriorates
Solution Approach 1:
The patent replaces traditional mechanical image processing methods with deep learning-based semantic segmentation models. These AI models automatically learn document boundaries and characteristics from training data, achieving superior segmentation performance without requiring manual tuning of processing parameters or complex rule-based systems.
Solution Approach 2:
The patent utilizes parameter changes in the form of learning rates, batch sizes, and model architectures to optimize segmentation performance. By adjusting these parameters during training and inference, the system achieves high accuracy in document segmentation while maintaining reasonable computational efficiency.
Data Source
AI summary
Systems and methods of the inventive subject matter are directed to artificial intelligence systems that are configured to receive an uploaded file from a user, where that upload contains images of one or more documents. Documents can be, e.g., receipts, identification cards, and so on. Once received, the system converts the image to a tensor to facilitate identifying mask images. The AI system then converts any identified mask images to grayscale before converting the image to black and white. White spaces are identified, and contours are identified that surround the white spaces. Next, white spaces with areas below a threshold size are discarded and white spaces with areas above that threshold have their contours stored. The AI system finds a minimum bounding rectangle for each contour and then crops the image around each white space accordingly. In this way, documents can be found within an image file and segmented to later processing.