Document Scanner Rectification Engine Perspective Distortion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional panorama stitching techniques struggle with capturing images of different orientations, angles, and depths, leading to distortion in the final stitched image, especially when applied to documents like magazines or posters.

Innovation Solution

An image processing system that uses a rectification engine to transform images with different perspectives and orientations onto a common plane, employing nearest neighbor clustering, run-length opening algorithms, and homography estimation to create a unified fronto-parallel view, allowing users to capture images at various angles without perfect alignment, and stitching them seamlessly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If conventional panorama stitching techniques are used to combine multiple images captured at different angles and orientations, then the ability to capture documents without perfect camera alignment is improved, but distortion in the final stitched image worsens

Engineering Contradiction:
Improvecamera alignment flexibilityVSAvoidimage distortion
Core Design Contradiction:
Ease of operationVSManufacturing precision

Solution Approach 1:

The patent applies preliminary rectification to individual images before stitching them together. The rectification engine processes each captured image to correct perspective distortion and align it with a canonical view of the document, ensuring that subsequent stitching operations work with pre-corrected images rather than raw distorted captures

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent segments the document capture process into two distinct stages: first, individual images are rectified independently to correct perspective distortion; second, the rectified images are stitched together. This segmentation allows each stage to optimize for its specific function without compromising the other

Inventive Principle:
Principle #1Segmentation

2Productivity

If multiple images are captured at random or predetermined patterns by multiple users or devices, then the flexibility and productivity of document scanning is improved, but the complexity of processing and aligning these diverse images worsens

Engineering Contradiction:
Improvedocument scanning efficiencyVSAvoidimage processing complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The rectification engine implements a universal processing pipeline that handles images from multiple sources (different users, devices, capture patterns) through the same rectification algorithms. This multi-functional approach processes diverse input images uniformly, eliminating the need for source-specific processing logic

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses feedback mechanisms where the rectification engine analyzes features from rectified images and adjusts processing parameters for subsequent images. The system learns from each processed image to improve alignment and rectification accuracy in real-time, handling increasing complexity through adaptive feedback

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If images are captured at different depths and orientations, then the versatility of the scanning system is improved, but the measurement precision of text recognition worsens

Engineering Contradiction:
Improvecapture angle flexibilityVSAvoidtext recognition accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The rectification engine performs preliminary geometric correction on each captured image to transform perspectives from different depths and orientations into a canonical fronto-parallel view. This pre-alignment ensures that text in rectified images appears as if captured from a perfect overhead angle, maximizing OCR accuracy before stitching occurs

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3284012B1Document scanner
Publication Date: 2021.12.29 GOOGLE LLC
  • EP3284012B1 patent drawingFigure 1
  • EP3284012B1 patent drawingFigure 2
  • EP3284012B1 patent drawingFigure 3

AI summary

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, to generate a scannable document. In one aspect, a method includes receiving a scan request, wherein the scan request includes a plurality of text images; for each text image of the plurality of text images: rectifying the text image to generate a text image with parallel image lines, generating a plurality of word bounding boxes that enclose one or more connected components in the text image, wherein each word bounding box is associated with a respective word, and generating, for each respective word in the text image, a plurality of points that represent the respective word; combining the plurality of text images to form a single text document; and providing the combined image as a scannable document.