Local Multiscale Tone Mapping for HDR Image Detail Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional tone-mapping operators for high dynamic range (HDR) images either lose significant image details, particularly in bright areas, or require high computational costs, making them inefficient for widespread use.
Innovation Solution
The implementation of a local multiscale tone mapping system that uses a combination of global and local tone mapping operators to generate tone-mapped images with improved local contrast preservation, employing a logarithmic domain processing and recursive mechanisms to maintain computational efficiency, allowing for efficient recovery of HDR images from tone-mapped metadata.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the Reinhard global photographic operator is used, then computational efficiency is improved, but image details are lost particularly in bright areas
Solution Approach 1:
The patent divides the image processing into multiple scales (different sigma values for Gaussian blurring) and processes different regions with different operators. The local operator processes regions with significant local contrast changes, while the global operator handles overall tone mapping, thereby preserving local details while maintaining computational efficiency.
Solution Approach 2:
The patent applies different tone mapping strategies to different regions of the image based on local characteristics. The local operator is applied where local contrast preservation is critical (edges, bright areas), while the global operator handles uniform regions, ensuring optimal detail preservation where needed without sacrificing overall efficiency.
2Loss of information
If the bilateral filter is used, then image details are preserved with high quality, but computational cost increases sixty times compared to simple TMOs
Solution Approach 1:
The patent segments the computation by identifying and processing only critical regions (where local contrast changes significantly) with the computationally intensive local operator, while applying the efficient global operator to the remaining regions. This selective approach preserves essential details without incurring the full computational cost of applying bilateral filtering to the entire image.
Solution Approach 2:
The patent applies the local operator selectively rather than universally, using it only where necessary to preserve local details. By controlling the application scope through scale parameter selection and regional analysis, the patent achieves sufficient detail preservation with fraction of the computational cost of full bilateral filtering.
3Loss of information
If local multiscale processing is implemented, then image details are minimized lost, but processing time increases compared to global tone mapping
Solution Approach 1:
The patent implements a multi-scale processing approach where computations are organized in a systematic sequence of scales (sigma values). By processing at multiple discrete scales and combining results, the patent achieves comprehensive detail preservation while maintaining structured, efficient computation that avoids redundant calculations.
Solution Approach 2:
The patent varies the scale parameter (sigma) systematically across different processing stages, allowing the same basic operator to adapt to different levels of detail preservation needs. This parameter-based approach enables flexible control over the trade-off between detail preservation and processing time without requiring fundamentally different algorithms at each scale.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
Techniques are provided to apply local multiscale tone mapping to relatively high dynamic range images. A global ratio operator may be applied to luminance values derived from an input high dynamic range (HDR) image to generate a high resolution gray scale ratio image comprising luminance ratio values. A local multiscale gray scale ratio image that comprises a weighted combination of at least two different gray scale ratio images may be generated. Each of the at least two different gray ratio images may be of a different spatial resolution level. Ratio values in the local multiscale gray scale ratio image may be used together with the luminance values derived from the input HDR image to generate a tone-mapped gray scale image.