Gain Map Image Compositing for HDR-SDR Luminance Fidelity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods fail to accurately superimpose objects on High Dynamic Range (HDR) images when converting them to Standard Dynamic Range (SDR) images, resulting in unexpected changes to the appearance of the superimposed elements due to differences in dynamic range and gain mapping.
Innovation Solution
An image processing apparatus and method that generates a second gain map based on the luminance of the superimposed object, updating the image file with both the original gain map and superimposed image data to ensure accurate conversion between HDR and SDR formats, using techniques like layering and partial gain maps to optimize file size and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an object is superimposed on an HDR image using a first gain map for HDR-to-SDR conversion, then the image can be converted to SDR format, but the superimposed object appears with unexpected changes in luminance and clarity
Solution Approach 1:
The patent divides the gain map application into two distinct segments: a first gain map for converting the HDR background image to SDR, and a second gain map specifically for converting the superimposed object to SDR. This segmentation allows each gain map to be optimized independently, preventing the luminance distortion that occurs when a single gain map is applied to both the background and foreground elements.
Solution Approach 2:
The patent applies different gain characteristics to different regions of the image based on their luminance properties. The second gain map is specifically tailored for the superimposed object's luminance range, ensuring that the object maintains its intended appearance. This local quality approach allows precise control over how different parts of the image are converted, preserving object fidelity while achieving overall HDR-to-SDR conversion.
2Device complexity
If a single gain map is used for both HDR image and superimposed object conversion, then processing is simplified, but the superimposed object does not appear as expected
Solution Approach 1:
The patent performs preliminary action by generating the second gain map in advance, based on the luminance characteristics of the superimposed object. This second gain map is prepared before the final composition, allowing the object to be pre-adjusted for accurate SDR display. By performing this preparation step beforehand, the system ensures reliable object appearance without requiring complex real-time adjustments during rendering.
3Adaptability or versatility
If HDR images are converted to SDR using existing gain maps, then display compatibility is improved, but superimposed objects lose their intended appearance
Solution Approach 1:
The patent creates a copy of the gain map concept specifically for the superimposed object. Instead of reusing the first gain map, a second gain map is generated that copies and adapts the gain principles to the object's specific luminance characteristics. This copying approach ensures that the object's visual information is preserved and accurately represented in the SDR output, separate from the background conversion process.
Data Source
AI summary
An image processing apparatus comprises: a superimposition unit that performs superimposition processing for superimposing an object on first image data of a first dynamic range, the first image data being associated with a first gain map for converting the first dynamic range into a second dynamic range, to generate superimposed image data; a generating unit that obtains a conversion characteristic for converting image data of the object into image data of the second dynamic range based on a luminance of the object, and generates a second gain map for converting the superimposed image data into second image data of the second dynamic range using the conversion characteristic; and an updating unit that updates an image file consisting of the first image data and the first gain map by using the superimposed image data and the second gain map.