Image Processing With Weighted Saturation-Aware Blur Sharpening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to effectively control the sharpness of blur and side effects caused by optical systems, particularly in luminance saturated areas, leading to issues like undershoot and ringing, which affect the accuracy of blur sharpening.
Innovation Solution
An image processing method that uses a machine learning model to generate a weight map based on optical system performance and luminance saturated areas, obtaining a weighted mean of the captured and estimated images to control sharpness and reduce side effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is used to sharpen blur in captured images, then the sharpness of blur is improved, but side effects such as undershoot and ringing occur in luminance saturated areas
Solution Approach 1:
The patent applies different processing strategies to different regions of the image based on luminance saturation. Luminance saturated areas are detected and handled differently from non-saturated areas, with the model output adjusted or suppressed in saturated regions to prevent side effects while maintaining sharpness in non-saturated regions.
Solution Approach 2:
The patent modifies the model output by changing parameters such as clipping values and scaling factors based on the detected luminance saturated areas. By adjusting these parameters dynamically according to local image characteristics, the system optimizes the balance between blur sharpness and side effect suppression.
2Measurement precision
If the sharpness of blur is increased to improve image quality, then measurement precision is improved, but harmful side effects increase
Solution Approach 1:
Different sharpness levels are applied to different regions of the image. Non-saturated areas receive full sharpening treatment for maximum sharpness, while saturated areas receive reduced or no sharpening to avoid side effects, creating a spatially varying quality profile.
Solution Approach 2:
The patent identifies luminance saturated areas where side effects would occur and uses this information to adjust the processing. By detecting and responding to the harmful conditions, the system converts the potential harm into an opportunity to optimize processing parameters and avoid side effects while maintaining quality where possible.
Data Source
Figure 1
Figure 2~3
Figure 4A~4C
AI summary
An image processing method includes generating a first image by inputting a captured image, which has been acquired by imaging using an optical system, into a machine learning model, acquiring information on optical performance of the optical system, and generating a second image based on the captured image, the first image, and first weight information. The first weight information is generated based on the information on the optical performance and information on a saturated area in the captured image.