AI HDR Image Denoising for Night Scene Detail Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for mobile terminals result in poor image quality at night scenes due to noise introduction during high dynamic range synthesis, leading to blurry and poorly defined images.
Innovation Solution
A method involving capturing multiple frames of original images, synthesizing them to create a high dynamic range (HDR) image, and applying artificial intelligence (AI)-based denoising to reduce noise and maintain image details, using a neural network model to identify and mitigate noise characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high dynamic range synthesis is performed on multiple frames of original images, then image dynamic range and detail are improved, but noise is introduced resulting in poor image definition
Solution Approach 1:
The patent segments the denoising process into multiple stages: first performing denoising on individual frames before synthesis, then performing denoising on the synthesized HDR image. This segmentation allows noise reduction at different processing levels, preventing noise accumulation while preserving dynamic range and detail information.
Solution Approach 2:
The patent applies preliminary denoising action by denoising individual frames before they are synthesized into the HDR image. This preliminary processing removes noise early in the pipeline, preventing it from propagating through subsequent synthesis operations and degrading final image quality.
2Manufacturing precision
If multiple frames of original images are captured for high dynamic range synthesis, then image quality is improved, but capture time is increased
Solution Approach 1:
The patent performs preliminary denoising on individual frames before synthesis, which allows using fewer frames while still achieving high image quality. By pre-processing frames to remove noise, the system can reduce the number of frames needed for synthesis, thereby decreasing total capture time while maintaining quality.
Solution Approach 2:
The patent changes the processing parameters by applying denoising operations that modify the characteristics of individual frames and the synthesized HDR image. These parameter changes enable better noise-rejection ratios with fewer frames, optimizing the balance between image quality and capture duration.
3Object-generated harmful factors
If denoising is performed on high luminance areas, then noise is reduced, but image details may be lost
Solution Approach 1:
The patent applies local quality by performing denoising operations specifically on high luminance areas where noise is most prominent, while preserving details in low luminance areas. The denoising algorithm selectively processes different regions with different strengths, maintaining local characteristics and avoiding over-smoothing that would lose detail.
Solution Approach 2:
The patent maintains continuity of useful action by performing denoising on both individual frames and the synthesized HDR image in sequence. This continuous denoising process preserves noise-rejection benefits across all processing stages while maintaining image details through careful algorithm design that distinguishes between noise and valid signal.
Data Source
Figure 1~3
Figure 4~6
Figure 7~9
AI summary
A method of image processing based on a plurality of frames of images and an electronic device. The method includes: capturing (101) a plurality of frames of original images; obtaining (102) a high dynamic range (HDR) image by performing image synthesis on the plurality of frames of original images; and performing (103) artificial intelligent-based denoising on the HDR image to obtain a target denoised image. By performing the method, image noise and effective details of the HDR image may be accurately distinguished, and an area with a high luminance level and an area with a low luminance level of the HDR image may be denoised, improving effectivity of denoising, such that the target denoised image obtained by denoising may have a denoised image and maintain image details at the same time, providing an optimal definition to the image.