SDR-to-HDR Conversion via User-Defined Composer Metadata
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for converting standard dynamic range (SDR) images to high dynamic range (HDR) images are inefficient and do not effectively preserve artistic intentions, especially when peak brightness exceeds 1000 nits.
Innovation Solution
The use of user-defined composer metadata generated through machine learning and user input, based on a dynamic SDR+ model template, to efficiently convert SDR images into HDR images with a user-defined look or appearance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing SDR-to-HDR conversion techniques are used, then conversion can be performed, but the conversion is inefficient and does not preserve artistic intentions, especially at peak brightness above 1000 nits
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models with extensive HDR image data and artistic intent annotations before actual conversion. The system prepares lookup tables, tone mapping curves, and style transfer models in advance, so that during runtime, the conversion process can efficiently apply pre-computed transformations while preserving artistic intentions without real-time computational overhead
Solution Approach 2:
The patent introduces an intermediary machine learning model that acts as a bridge between SDR input and HDR output. This intermediary model, trained on diverse HDR content with artistic annotations, translates SDR images to HDR while maintaining artistic intentions. The model serves as a mediator that understands both the technical conversion requirements and the artistic preservation goals
2Manufacturing precision
If user-defined composer metadata is generated through machine learning and user input, then high-quality HDR images with artistic intentions can be achieved, but the process complexity increases
Solution Approach 1:
The patent applies self-service by enabling users to define their own composer metadata and artistic preferences directly in the system. Users can adjust tone mapping parameters, brightness levels, and stylistic properties through intuitive interfaces, and the system automatically processes these inputs to generate optimized HDR output without requiring external expert intervention or complex manual configuration
Solution Approach 2:
The patent utilizes parameter changes by allowing dynamic adjustment of conversion parameters such as peak brightness, tone mapping curves, and color space transformations. The machine learning model accepts user-defined parameters and automatically optimizes the conversion process based on these inputs, enabling flexible control over HDR image quality while managing complexity through parameterized configurations
3Reliability
If conventional tone mapping is used, then SDR images can be converted to HDR, but the conversion quality degrades at peak brightness levels above 1000 nits
Solution Approach 1:
The patent applies dynamics by implementing adaptive tone mapping that dynamically adjusts conversion parameters based on the input image characteristics and target display capabilities. The machine learning model dynamically selects appropriate tone mapping strategies, adjusts brightness distribution, and optimizes color mapping in real-time based on the specific content being converted, ensuring high quality across varying peak brightness levels including above 1000 nits
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
Backward reshaping metadata prediction models are trained with training SDR images and corresponding training HDR images. Content creation user input to define user adjusted HDR appearances for the corresponding training HDR images is received. Content-creation-user-specific modified backward reshaping metadata prediction models are generated based on the trained prediction models and the content creation user input. The content-creation-user-specific modified prediction models are used to predict operational parameter values of content-creation-user-specific backward reshaping mappings for backward reshaping SDR images into mapped HDR images of at least one content-creation-user-adjusted HDR appearance.