NIR-Guided Visible Image Denoising via Gradient Scale Map

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image de-noising methods, especially under low-light conditions, struggle to distinguish between noise and original image signals, leading to over-smoothened results with artifacts, especially when noise levels are high, and often require intrusive or environmentally unsuitable lighting sources.

Innovation Solution

A multispectral imaging system that captures both visible light and near-infrared (NIR) images of the same scene, aligns them pixel-wise, and generates a gradient scale map to guide the de-noising process, using the NIR image to enhance the quality of the visible light image without introducing additional artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If high ISO is used to increase signal strength in low light conditions, then image brightness is improved, but image noise increases

Engineering Contradiction:
Improveimage brightnessVSAvoidimage noise
Core Design Contradiction:
Illumination intensityVSObject-generated harmful factors

Solution Approach 1:

The patent introduces a gradient scale map as an intermediary between the noisy visible light image and the denoised output. This gradient scale map, derived from comparing gradient fields of multiple images, acts as a mediator that guides the denoising process to preserve edges and structures while removing noise, thus resolving the contradiction between brightness and noise

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent combines information from multiple images (visible light images and near-infrared images) to create a composite gradient scale map. This composite approach integrates multiple sources of information to achieve better denoising performance while preserving image quality, effectively addressing the noise-brightness tradeoff

Inventive Principle:
Principle #40Composite materials

2Productivity

If conventional single-image de-noising methods are applied to high noise level images, then processing speed is maintained, but image quality deteriorates with over-smoothening and artifacts

Engineering Contradiction:
Improveprocessing speedVSAvoidimage quality
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent transitions from single-image processing to multi-image processing by incorporating near-infrared images alongside visible light images. This dimensional expansion allows the gradient scale map to capture structural information that is not available in single visible light images, enabling effective denoising without over-smoothening while maintaining computational efficiency

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

3Manufacturing precision

If visible flash is used as guidance image for de-noising, then image quality is improved, but intrusive lighting effects and artifacts are introduced

Engineering Contradiction:
Improveimage qualityVSAvoidartifacts and intrusive lighting
Core Design Contradiction:
Manufacturing precisionVSObject-generated harmful factors

Solution Approach 1:

The patent changes the wavelength parameter of the guidance light from visible spectrum to near-infrared spectrum. This parameter change allows the guidance image to be captured outside the visible range, eliminating intrusive lighting effects and artifacts while still providing structural information for denoising the visible light image

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2984621B1Near infrared guided image denoising
Publication Date: 2019.05.15 QUALCOMM INC
  • EP2984621B1 patent drawingFigure 1
  • EP2984621B1 patent drawingFigure 2
  • EP2984621B1 patent drawingFigure 3A~3B

AI summary

Systems and methods for multispectral imaging are disclosed. The multispectral imaging system can include a near infrared (NIR) imaging sensor and a visible imaging sensor. The disclosed systems and methods can be implemented to de-noise a visible light image using a gradient scale map generated from gradient vectors in the visible light image and a NIR image. The gradient scale map may be used to determine the amount of de-noising guidance applied from the NIR image to the visible light image on a pixel-by-pixel basis.