Document Occlusion Artifact Removal via Depth Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Physical documents scanned into digital form often contain occlusion artifacts from objects obstructing the view, which degrade image quality and hinder optical character recognition (OCR) performance.
Innovation Solution
A method that determines depth information for each pixel in a document image to identify and remove occlusion artifacts by setting pixels with greater depth than the baseline document depth to a minimum value and using inpainting techniques to reconstruct the image, while adjusting document boundaries for improved alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If depth-based artifact removal is applied, then image quality and OCR performance are improved, but processing complexity increases
Solution Approach 1:
The system performs depth estimation and artifact identification before the main OCR processing step. By预先 identifying and removing occlusion artifacts based on depth information, the subsequent OCR process operates on cleaner data, improving overall reliability while managing complexity through staged processing
Solution Approach 2:
Depth information serves as an intermediary layer between the raw image and the final OCR processing. The depth map acts as a mediator that identifies artifact locations without requiring direct modification of the OCR engine, thus improving OCR performance while isolating the complexity to a separate depth estimation module
2Measurement precision
If depth information is calculated for each pixel, then artifact identification accuracy is improved, but computational load increases
Solution Approach 1:
The system calculates depth information for all pixels (excessive action) to ensure complete artifact detection, but applies this depth information selectively only where needed for artifact removal. This approach guarantees high identification accuracy while managing computational load by avoiding unnecessary processing in artifact-free regions
Solution Approach 2:
The system transforms the image from standard RGB color space to depth map representation, changing the parameter domain to enable artifact detection. This parameter transformation allows accurate depth-based identification without requiring computationally intensive analysis in the original color space
Data Source
AI summary
Depth information for each pixel of a digitally captured image of a document occluded by an object is determined. The pixels the pixels for which the depth information is indicative of a greater depth than a baseline depth of the document are identified as belonging to an artifact corresponding to the object occluding the document. The artifact is removed from the digitally captured image.