Adaptive SDR-to-HDR Luminance Mapping with Pixel Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for converting Standard-Dynamic Range (SDR) images to High-Dynamic Range (HDR) images face challenges such as lack of self-adaptability to image content, complex calculations, and excessive dynamic range expansion leading to artifacts like banded false contours.
Innovation Solution
A method involving luminance statistical analysis, pixel clustering based on luminance flag values, and generation of a conversion mapping relationship using Bessel curves or lookup tables to transform SDR images into HDR images, with adaptive adjustments for scene changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing methods convert SDR images to HDR images using traditional algorithms, then the conversion can be performed, but the methods lack self-adaptability to different image contents and produce artifacts like banded false contours
Solution Approach 1:
The patent changes the parameter of luminance flag values dynamically based on image content statistics. By calculating the standard deviation of luminance values and adjusting the flag value threshold accordingly, the method adapts to different image contents (high-frequency vs. low-frequency images) and prevents artifacts like banded false contours while maintaining conversion accuracy.
2Measurement precision
If existing methods use complex conversion algorithms to achieve accurate SDR to HDR conversion, then conversion accuracy can be improved, but the computational complexity increases
Solution Approach 1:
The patent segments the luminance value range into multiple intervals using luminance flag values. By dividing the conversion process into discrete segments based on luminance thresholds, the method simplifies the computational complexity while maintaining conversion accuracy. Each segment can be processed independently using simple lookup tables or linear transformations.
Solution Approach 2:
The method uses self-adaptive threshold selection where the luminance flag values are automatically determined from the image's own luminance statistics (standard deviation). This eliminates the need for complex external parameter tuning or iterative optimization algorithms, reducing computational complexity while achieving accurate conversion.
3Illumination intensity
If the dynamic range expansion is excessive during SDR to HDR conversion, then the HDR effect is enhanced, but artifacts like banded false contours appear
Solution Approach 1:
The patent dynamically adjusts the luminance flag value threshold based on the image's luminance standard deviation. For high-frequency images with large standard deviation, higher flag values are used to prevent excessive expansion in detailed regions. For low-frequency images with small standard deviation, lower flag values are used to maintain smooth gradients. This adaptive parameter adjustment prevents banded false contours while preserving HDR effects.
Data Source
AI summary
This application relates to a method and a device for changing a dynamic range of luminance of an image. The method comprises: extracting luminance statistical information of at least a portion of pixels of a frame to be processed, wherein the frame to be processed has an original dynamic range of luminance; clustering the at least a portion of pixels into pixel groups divided by multiple luminance flag values based on the luminance statistical information of at least a portion of pixels, wherein each of the pixel groups includes a predetermined number of pixels, and pixels of each two adjacent pixel groups are divided by one of the multiple luminance flag values; and generating a conversion mapping relationship based on the multiple luminance flag values such that the frame to be processed is converted into a target frame with a target dynamic range of luminance based on the conversion mapping relationship, wherein the conversion mapping relationship associates original luminance values in the original dynamic range of luminance with target luminance values in the target dynamic range of luminance.