Darkness Gamma Transfer Function for HDR Image Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High dynamic range (HDR) images compressed to low dynamic range (LDR) for display on LDR devices often result in reduced contrast and loss of details, particularly in dark areas, due to existing tone mapping techniques that can produce unnatural shadows and artifacts.
Innovation Solution
The use of a darkness gamma transfer function to calculate gain values for each output pixel, allowing for a universal transfer curve that preserves natural image details without over-contrast in dark areas, and is computationally efficient with a simple iterative process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If regular gamma-correction is used for HDR to LDR compression, then the output appears more natural and pleasant, but the differences between closest input values become too large causing overexposed shadows
Solution Approach 1:
The patent modifies the gamma correction function by introducing a new parameter alpha to control the transition region. The function changes form based on whether input values are below or above a threshold, allowing precise control over how shadow details are preserved while maintaining natural appearance in other regions.
Solution Approach 2:
The transfer function is divided into two segments: one for values below the threshold (preserving shadow details with different characteristics) and one for values above the threshold (maintaining natural gamma correction). This segmentation allows different processing strategies for different parts of the image.
2Stability of the object's composition
If logarithmic correction is used for strong HDR cases, then the transformation is smooth, but the contrast is reduced and less details are preserved
Solution Approach 1:
The patent introduces a threshold parameter and an alpha parameter to control the behavior of the transfer function. By changing the function form based on the input value relative to the threshold, it avoids the continuous compression of logarithmic functions while maintaining smooth transitions.
Solution Approach 2:
The transfer function dynamically adapts its characteristics based on the input value. For shadow regions (below threshold), it applies one transformation; for other regions, it applies a different transformation. This dynamic behavior preserves details while maintaining smooth overall appearance.
3Illumination intensity
If global tone mapping is applied to preserve naturalness, then dark area details are maintained, but additional local corrections are affected and artifacts may appear
Solution Approach 1:
The patent modifies the global tone mapping function by introducing a threshold parameter and alpha parameter. This creates a transfer function that naturally handles dark areas without creating discontinuities in the derivative, thereby preserving details while avoiding the artifacts that would result from subsequent local corrections.
4Device complexity
If linear scaling is used for HDR to LDR compression, then the process is simple, but the resulting image is very dark requiring brightness fixation that loses information
Solution Approach 1:
The patent enhances simple linear scaling by introducing a piecewise function with a threshold parameter. This maintains the simplicity of the approach while adding the capability to preserve dark area details through a different transformation in the shadow region, avoiding information loss.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An example apparatus for compressing dynamic range includes an image receiver to receive an input image with a high dynamic range. The apparatus further includes a darkness gamma transfer calculator to calculate gain values for each output pixel via a darkness gamma transfer function. The apparatus also further includes a gain applicator to apply the gain values to color channel values of the input image to generate a compressed image.