Neural Tone Mapping for SDR-to-HDR Display Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Legacy content in standard dynamic range (SDR) cannot be efficiently converted to high-dynamic range (HDR) for display due to limitations in metadata availability and computational constraints, hindering the full utilization of HDR display capabilities.
Innovation Solution
A neural network-based approach is employed to predict HDR statistics and metadata from SDR images, utilizing two neural networks for global and local tone mapping to generate optimized tone-mapping curves, enabling effective conversion and display management of SDR images on HDR displays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If neural networks are used to predict HDR statistics and generate tone-mapping curves for SDR images, then dynamic range conversion capability is improved, but computational complexity and processing time increase
Solution Approach 1:
The neural network models are pre-trained offline using pairs of SDR and HDR images to learn the mapping relationships. During actual deployment, only the trained model parameters are applied to convert SDR images to HDR, avoiding the need for complex real-time training computations while maintaining high conversion accuracy
Solution Approach 2:
Instead of performing complex real-time computations to convert SDR to HDR, the system uses pre-computed neural network models that have learned the conversion relationships from training data. The model predicts HDR statistics and generates tone-mapping curves by copying the learned patterns from training examples, significantly reducing runtime computational complexity
2Measurement precision
If metadata transmission is used for SDR to HDR conversion, then conversion accuracy is improved, but transmission bandwidth requirements increase
Solution Approach 1:
The system extracts and processes only the essential input image data through neural network inference to generate HDR statistics and tone-mapping curves. By removing the dependency on extensive metadata transmission and using the neural network to infer necessary information from the image content itself, the system achieves accurate conversion with minimal transmission requirements
Solution Approach 2:
The neural network model acts as an intermediary that processes SDR image data and transforms it into HDR representations. Instead of relying on direct metadata transmission for conversion, the neural network serves as an intelligent mediator that learns the mapping relationships from training data and applies them to convert images accurately without requiring extensive metadata
3Adaptability or versatility
If tone-mapping curves are generated using predicted HDR statistics, then display management capability is improved, but processing time increases
Solution Approach 1:
The neural network performs preliminary prediction of HDR statistics from SDR images, which then enables efficient tone-mapping curve generation. By pre-computing the statistical characteristics and mapping relationships through the neural network, the subsequent tone-mapping process becomes faster and more efficient, improving overall display management capability while controlling processing time
Data Source
AI summary
Methods and systems for dynamic range conversion and display mapping of standard dynamic range (SDR) images onto high dynamic range (HDR) displays are described. Given an SDR input image, a processor generates an intensity (luminance) image and optionally a base layer image and a detail layer image. A first neural network uses the intensity image to predict statistics of the SDR image in a higher dynamic range. These predicted statistics together with the original image statistics of the input image are used to derive an optimal tone-mapping curve to map the input SDR image onto an HDR display. Optionally, a second neural network, using the intensity image and the detail layer image, can generate a residual detail layer image in a higher dynamic range to enhance the tone-mapping of the base layer image into the higher dynamic range.