Anisotropic Diffusion for Image Relighting Edge Preservation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional image relighting techniques struggle to effectively remove unwanted illumination, particularly spurious edges, which can mask real edges and reduce the aesthetic quality of images, and often cause global intensity variations that are undesirable.

Innovation Solution

The method involves performing anisotropic diffusion on digital images to separate reflectance and illumination components, using either conventional or model-based anisotropic diffusion to estimate and remove the original illumination, and then applying a new illumination to the reflectance image, allowing for natural and desirable image relighting.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If conventional illumination normalization techniques (level compression, histogram stretching) are used to reduce illumination variations, then the perception of illumination gradients is reduced, but sharp spurious edges cannot be properly compensated and global intensity variations are introduced

Engineering Contradiction:

Improveillumination variationsVSAvoidspurious edge compensation

Core Design Contradiction:

Illumination intensityVSManufacturing precision

Solution Approach 1:

The patent segments the image processing into distinct components: reflectance estimation (separating object properties from illumination) and illumination normalization. This segmentation allows targeted handling of spurious edges through anisotropic diffusion applied specifically to the reflectance map, while global illumination variations are corrected separately through histogram matching, avoiding the trade-off present in conventional unified approaches

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by using anisotropic diffusion with directionally selective smoothing that preserves edges while removing spurious artifacts. The diffusion process adapts locally to the image structure, applying different smoothing strengths in different directions and locations, thereby maintaining sharp real edges while eliminating unwanted illumination-induced spurious edges

Inventive Principle:

Principle #3Local quality

2Loss of information

If conventional illumination normalization techniques are applied to remove unwanted illumination, then overall contrast is improved, but spurious edges are not properly removed and may be masked

Engineering Contradiction:

Improveinformation preservationVSAvoidspurious edges

Core Design Contradiction:

Loss of informationVSObject-generated harmful factors

Solution Approach 1:

The patent inverts the conventional approach by first estimating the reflectance map through anisotropic diffusion (removing illumination effects), then applying illumination normalization to the reflectance rather than the original image. This inversion ensures that spurious edges are removed during the reflectance estimation phase before normalization occurs, preventing their propagation through the processing pipeline

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The reflectance map serves as an intermediary representation that separates object properties from illumination effects. By processing through this intermediate reflectance map rather than directly on the illuminated image, the patent enables selective removal of spurious edges while preserving real structural information, which is then combined with normalized illumination for the final result

Inventive Principle:

Principle #24Intermediary (Mediator)

3Manufacturing precision

If anisotropic diffusion is performed to separate reflectance and illumination components, then spurious edges are removed and real edges are preserved, but processing complexity increases

Engineering Contradiction:

Improveedge preservationVSAvoidprocessing complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent manages processing complexity through parameter optimization in the anisotropic diffusion process. By carefully selecting diffusion coefficients, iteration counts, and kernel sizes, the implementation achieves effective reflectance estimation and spurious edge removal while maintaining computational efficiency. The parameters are tuned to balance processing complexity against the quality of edge preservation and spurious edge elimination

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS7295716B1Method and apparatus for diffusion based image relighting

Publication Date: 2007.11.13 FUJIFILM CORP

AI summary

A method for image relighting is presented which receives an input image having at least one spurious edge directly resulting from a first illumination present when the input image was acquired, performs anisotropic diffusion on the input image to form a diffusion image, removes the first illumination using the diffusion image to generate a reflectance image and applies a second illumination to the reflectance image. An apparatus for relighting is presented which includes a processor operably coupled to memory storing input image having a first illumination present when the input image was acquired, and functional processing including an anisotropic diffusion module to perform anisotropic diffusion on the input image to form a diffusion image, a combination module which removes the first illumination using the diffusion image to generate a reflectance image, a second illumination module which generates a second illumination, and a lighting application model which applies the second illumination.