Haze Removal in High-Luminance Image Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques, such as the dark channel prior method, struggle to effectively remove haze from high-luminance image regions without losing detail, as they incorrectly estimate the haze effect leading to saturation and loss of fine luminance and color details.
Innovation Solution
An image processing device and method that calculates a superimposed pixel value and effect estimation value based on the input image and transmittance, allowing for targeted correction of haze effects in high-luminance regions to prevent detail loss, by adjusting pixel values using a transmittance value that decreases from 1 to 0, and performing corrections based on specific limit values and boundary values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If the dark channel prior method is applied to remove haze from high-luminance regions, then haze removal is achieved, but detail information (fine luminance and color changes) is lost due to saturation
Solution Approach 1:
The patent applies different correction strategies to different regions of the image based on local characteristics. High-luminance regions are identified and processed differently from other regions, using adjusted transmittance values and correction amounts to preserve local detail while removing haze. This local differentiation prevents the saturation problem that occurs when uniform correction is applied across the entire image.
Solution Approach 2:
The patent modifies the transmittance parameter t(x) and the correction amount parameter based on the effect estimation value. By changing these parameters dynamically according to local image characteristics, the correction process adapts to preserve detail in high-luminance regions while effectively removing haze. The transmittance is adjusted to be less than 0 in certain regions, and correction amounts are modulated based on effect estimation values.
2Object-affected harmful factors
If correction is applied to high-luminance regions with the assumption that haze effect is large, then haze removal is performed, but the correction amount becomes excessive causing saturation
Solution Approach 1:
The patent introduces an effect estimation value that provides feedback on the actual haze effect in each region. This feedback mechanism allows the correction process to adjust the correction amount dynamically, preventing excessive correction in high-luminance regions. The effect estimation value guides the correction amount modulation, ensuring precision by adapting to local conditions rather than applying uniform correction.
Solution Approach 2:
The patent applies partial correction to high-luminance regions by modulating the correction amount based on effect estimation values. Instead of applying full correction uniformly, the system applies only the necessary correction amount in each region, preventing saturation while still removing haze effectively. This partial action approach maintains correction precision.
Data Source
AI summary
The present invention provides an image processing device and an image processing method capable of removing haze (reducing the effect of haze) without losing detail of a high-luminance subject. In the aspect of the present invention, the input image I is represented by I=J·t+A·(1−t) where an original image is J, an atmospheric light pixel value is A, and a transmittance is t. In this case, a dark channel value D of each pixel of the input image I is calculated, and is associated with the transmittance t having a value monotonically decreasing for each pixel of which D ranges from 0 to 1 and associated with the transmittance t having a value monotonically increasing for each pixel which corresponds to haze and of which D ranges from 1 to Dmax. In such a manner, the original image J is generated as a corrected image.