HDR Image Enhancement With Neural Super-Resolution Color Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image enhancement methods struggle to maintain consistency in color, brightness, contrast, and saturation when performing real-time super-resolution processing on high-resolution images, particularly in high dynamic range (HDR) environments, leading to suboptimal image quality.
Innovation Solution
An image enhancement method utilizing a neural network model that incorporates color, texture, and multi-scale attention mechanisms to perform super-resolution processing on high dynamic range images, ensuring consistent enhancement of these attributes across different regions and scales.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If downsampling + HDR + super-resolution processing is used to meet real-time requirements, then processing speed is improved, but image quality consistency (color, brightness, contrast, saturation) deteriorates
Solution Approach 1:
The patent segments the image processing into distinct stages: downsampling to low resolution, HDR enhancement, super-resolution restoration, and color correction. Each stage handles specific aspects of image enhancement, allowing optimization at each step while maintaining overall image quality consistency.
Solution Approach 2:
The patent applies local quality adjustment through color correction maps that are generated based on the enhanced image. These maps apply localized adjustments to color, brightness, contrast, and saturation in different regions of the image, ensuring consistency across the entire image while allowing regional optimizations.
2Use of energy by moving object
If complex super-resolution is performed only on luminance channel, then computing overhead is reduced, but color and brightness consistency deteriorates
Solution Approach 1:
The patent introduces color correction maps as an intermediary element between the luminance-based super-resolution output and the final color channel. These maps serve as mediators that transfer and adjust color information, ensuring that color and brightness consistency is maintained without requiring complex super-resolution processing on the color channels themselves.
Solution Approach 2:
The patent changes the processing parameters by applying different operations to different channels: complex super-resolution is applied to the luminance channel while simpler upsampling is used for color channels, with subsequent color correction applied via generated correction maps. This parameter differentiation reduces computing overhead while maintaining color and brightness consistency.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
An image enhancement method and apparatus are disclosed. The image enhancement method includes: obtaining a first high dynamic range HDR image corresponding to a to-be-processed image and a color image feature of the to-be-processed image, wherein the color image feature indicates different luminance regions or different color change regions in the to-be-processed image, the to-be-processed image is an image of a first resolution, the first HDR image is an image of a second resolution, and the first resolution is higher than the second resolution (610); inputting the first HDR image into a neural network model for performing super-resolution processing (620); and performing, by using the neural network model, image enhancement processing on the super-resolution processed first HDR image and the color image feature, to obtain a second HDR image corresponding to the to-be-processed image, wherein the second HDR image is an HDR image whose resolution is the first resolution (630). The method can improve image enhancement processing effect.