Semantic Tone Mapping for HDR Images in High-Noise Scenes

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing techniques in mobile devices often result in undesirable artifacts such as darkened important areas and creation of brighter or darker spots within objects, leading to decreased image quality and user satisfaction, particularly in high noise and high dynamic range environments.

Innovation Solution

A method involving machine learning segmentation-based tone mapping that generates a high dynamic range blended image and performs a tone fusion operation using a semantic delta-weight map, synthesizing low dynamic range images and generating filtered weight maps to produce a fused image, respecting object boundaries and reducing artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If existing image processing techniques are used, then processing speed is maintained, but image quality deteriorates due to darkening of important areas and creation of brighter or darker spots within objects

Engineering Contradiction:

Improveimage qualityVSAvoidartifacts

Core Design Contradiction:

Manufacturing precisionVSObject-generated harmful factors

Solution Approach 1:

The patent applies semantic segmentation to divide the image into different semantic regions (e.g., sky, ground, objects). This segmentation is then used to generate delta-weight maps that guide the tone mapping process, allowing different regions to be processed differently. This resolves the contradiction by preventing uniform tone mapping that causes artifacts in important areas while maintaining processing efficiency through region-based processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements local tone mapping by generating delta-weight maps specific to each semantic region. Instead of applying global tone mapping, the system adjusts tone parameters locally based on the semantic content and noise characteristics of each region. This allows important areas to be protected from darkening while maintaining overall image quality, resolving the contradiction between processing speed and image quality.

Inventive Principle:

Principle #3Local quality

2Manufacturing precision

If tone mapping is applied without semantic segmentation, then processing complexity is reduced, but image quality deteriorates due to unwanted spots and darkening artifacts

Engineering Contradiction:

Improveimage qualityVSAvoidprocessing complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent uses semantic segmentation to classify pixels into different semantic regions, which then guides the tone mapping process. The segmentation results are used to generate delta-weight maps that control the tone mapping parameters for each region. This approach increases processing complexity only moderately while significantly improving image quality by preventing artifacts in important areas.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces delta-weight maps as an intermediary between the semantic segmentation results and the tone mapping operation. These weight maps serve as a mediator that translates semantic information into tone mapping controls, allowing the system to achieve high image quality without directly implementing complex region-based processing at every stage.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Manufacturing precision

If traditional tone mapping is used in high noise environments, then processing speed is maintained, but image quality deteriorates due to noise amplification and artifacts

Engineering Contradiction:

Improveimage qualityVSAvoidnoise

Core Design Contradiction:

Manufacturing precisionVSObject-affected harmful factors

Solution Approach 1:

The patent applies local tone mapping that adapts to the noise characteristics of each semantic region. By analyzing noise patterns within each region and adjusting tone mapping parameters accordingly, the system prevents noise amplification in important areas while maintaining processing speed through region-based processing rather than pixel-by-pixel analysis.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent converts the harmful noise present in high noise environments into useful information for guiding the tone mapping process. By analyzing noise patterns and using them to adjust delta-weight maps, the system leverages noise characteristics to improve tone mapping accuracy and prevent artifacts, turning the harmful noise into a beneficial guide for region-based processing.

Inventive Principle:

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

Data Source

PatentUS12518364B2Machine learning segmentation-based tone mapping in high noise and high dynamic range environments or other environments

Publication Date: 2026.01.06 SAMSUNG ELECTRONICS CO LTD

AI summary

A method includes obtaining multiple input image frames and generating a high dynamic range (HDR) blended image based on the input image frames, where the HDR blended image has a higher dynamic range than individual ones of the input image frames. The method also includes performing a tone fusion operation on the HDR blended image based on a semantic delta-weight map to generate a fused image. Performing the tone fusion operation includes synthesizing multiple low dynamic range (LDR) images based on the HDR blended image and generating initial weight maps based on the LDR images. Performing the tone fusion operation also includes generating filtered weight maps based on the initial weight maps, the semantic delta-weight map, and a guided filter. Performing the tone fusion operation further includes generating the fused image based on the filtered weight maps and decomposed versions of the LDR images.