Fisheye Lens Distortion Correction via 3D Mapping

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Omnidirectional cameras capture 360-degree video with fisheye lenses, resulting in distorted 2-dimensional images that need to be mapped to a 3-dimensional representation for accurate 360-degree video presentation, as the images are warped and compressed, lacking realistic distortion correction.

Innovation Solution

A method and apparatus that process 2-dimensional video data from omnidirectional cameras by mapping pixels in a circular image region to corresponding locations on a 3-dimensional hemispherical representation using polynomial equations, adjusting radial values for non-linear warping, and incorporating scaling parameters to correct distortion, enabling accurate 360-degree video presentation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If fisheye lenses are used to capture 360-degree video with omnidirectional cameras, then a wide field of view can be captured, but the resulting images exhibit severe distortion and warping

Engineering Contradiction:
Improvefield of viewVSAvoidimage distortion
Core Design Contradiction:
Area of stationary objectVSManufacturing precision

Solution Approach 1:

The patent transforms the distorted 2D fisheye image coordinates into a 3D spherical coordinate system, then maps them to corrected 2D output coordinates. This dimensional transformation allows the system to recover the true geometric relationships of the captured scene by introducing an intermediate 3D spherical representation that preserves angular information while eliminating radial distortion.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent applies parameter changes by using polynomial equations with adjustable coefficients to model the non-linear distortion characteristics of fisheye lenses. By fitting these polynomials to calibration data and adjusting the parameters during the coordinate transformation process, the system can accurately correct various types of fisheye distortion while maintaining the wide field of view.

Inventive Principle:
Principle #35Parameter changes

2Quantity of substance

If 2-dimensional images are captured to represent 360-degree scenes, then data storage and transmission are simplified, but the scenes appear distorted and lack realistic representation

Engineering Contradiction:
Improvedata volumeVSAvoidscene representation accuracy
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent introduces a virtual 3D spherical coordinate system as an intermediate representation layer between the captured 2D fisheye images and the final corrected output. This intermediate 3D representation preserves the angular and spatial relationships of the original scene while allowing for accurate mapping back to 2D display coordinates, thereby maintaining scene representation accuracy without increasing data volume.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Measurement precision

If polynomial equations with adjusted radial values are used to map pixels, then distortion correction accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improvedistortion correction accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary action by pre-calculating and storing polynomial coefficients through a calibration process using known reference patterns. These pre-computed parameters are then reused during real-time distortion correction, significantly reducing the computational burden during actual video processing while maintaining high correction accuracy. The heavy computational work is shifted to an offline calibration stage.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3465607B1Fisheye rendering with lens distortion correction for 360-degree video
Publication Date: 2021.02.17 QUALCOMM INC
  • EP3465607B1 patent drawingFigure 1
  • EP3465607B1 patent drawingFigure 2A~2B
  • EP3465607B1 patent drawingFigure 3

AI summary

In various implementations, provided are systems and methods for correcting the distortion present in a fisheye image, and rendering the image for display as 360-degree video. In various implementations, a computing device can receive 2-dimensional video data captured by an omnidirectional camera. The computing device can map an image from each video frame to a 3-dimensional hemispherical representation. In various implementations, this mapping can be executed using a polynomial model. The 3-dimensional hemispherical representation can then be used in a 360-degree video presentation, to provide a virtual reality experience.