SDR-to-HDR Local Reshaping With Co-Occurrence Denoising

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing techniques struggle to convert Standard Dynamic Range (SDR) images to High Dynamic Range (HDR) images without introducing visually noticeable noise-related artifacts, especially on big screen displays where the Human Visual System can resolve more details.

Innovation Solution

Implement denoising techniques in conjunction with reshaping operations to generate HDR images from SDR images, using local reshaping function index maps denoised by co-occurrence filtering to reduce noise and enhance visual quality, preserving local contrast and color saturation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If SDR images are converted to HDR images using reshaping operations, then visual quality and dynamic range are improved, but noise-induced artifacts become more noticeable

Engineering Contradiction:
Improvedynamic rangeVSAvoidnoise-induced artifacts
Core Design Contradiction:
Illumination intensityVSObject-affected harmful factors

Solution Approach 1:

The patent applies denoising operations to the SDR image before performing reshaping operations to convert to HDR. This preliminary denoising step removes noise from the source image, preventing it from being amplified during the subsequent dynamic range expansion, thereby achieving both improved dynamic range and reduced noise artifacts in the final HDR image

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent identifies that noise in SDR images becomes problematic during HDR conversion, and transforms this harmful effect by applying targeted denoising processing. The denoising operation converts the problematic noise into benefit by removing it before it can be amplified, while preserving important image details and edges through intelligent noise discrimination

Inventive Principle:
Principle #22Blessing in disguise (Convert harm into benefit)

2Object-affected harmful factors

If denoising operations are applied to SDR images before reshaping, then noise-induced artifacts are reduced, but image processing complexity increases

Engineering Contradiction:
Improvenoise-induced artifactsVSAvoidimage processing complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary denoising processing stage between the SDR image input and the HDR reshaping operation. This intermediary step acts as a mediator that cleans the input image before it undergoes dynamic range expansion, effectively reducing noise artifacts in the final HDR output while maintaining a structured and manageable processing pipeline

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent modifies the processing parameters by applying denoising filters with specific characteristics to the SDR image before reshaping. By changing the state of the image through denoising parameters (such as filter strength, kernel size, or denoising algorithm parameters), the system reduces noise-induced artifacts while keeping the processing complexity controlled through parameter optimization

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4441697B1Denoising for SDR-to-HDR local reshaping
Publication Date: 2025.12.24 DOLBY LABORATORIES LICENSING CORP
  • EP4441697B1 patent drawingFigure 1
  • EP4441697B1 patent drawingFigure 2A
  • EP4441697B1 patent drawingFigure 2B~2C

AI summary

An example approach relates to generating images of a first dynamic range from input images of a second dynamic range lower than the first dynamic range. A first array of co-occurrence statistics in connection with an input image of the second dynamic range is generated. A second array of reshaping function index values for selecting reshaping functions from among a plurality of candidate reshaping functions is generated. To generate a reshaped image of the first dynamic range, denoising operations are performed based at least in part on the first array of co-occurrence statistics, whereas reshaping operations are performed based at least in part on the second array of reshaping function index values.