Pixel-Based Gray Image Dehazing Without Patch Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional haze removal algorithms using the dark channel prior require large local patches, leading to high computational costs and processing times, and often result in blocked artifacts and inaccurate atmospheric light estimation, especially in images with low-frequency textures or bright objects.
Innovation Solution
A pixel-based method that generates a blended gray image by combining color channels, calculates atmospheric light, and uses single pixel operations to enhance image contrast and visibility without soft-matting, thereby reducing processing time and artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional haze removal algorithms use dark channel prior with large local patches, then haze removal can be achieved, but computational costs increase and processing time becomes excessively long
Solution Approach 1:
The patent segments the image processing task into pixel-level operations rather than using large local patches. Each pixel is processed independently using single-pixel operations to estimate atmospheric light and generate transmission maps, eliminating the need for computationally intensive patch-based dark channel prior calculations while maintaining haze removal effectiveness
Solution Approach 2:
The patent replaces the mechanical/computational system of sliding window patch processing with a pixel-based mathematical model. By substituting the physical patch-based dark channel prior calculation with a direct pixel-level atmospheric light estimation model, the system achieves the same haze removal function with significantly reduced computational complexity
2Manufacturing precision
If conventional algorithms use additional refinement steps and soft matting, then image quality may be improved, but device complexity increases and processing time increases
Solution Approach 1:
The patent extracts and removes the unnecessary refinement steps and soft matting operations from the conventional algorithm pipeline. By taking out these additional processing stages, the system achieves the same image quality through a simplified direct pixel-based approach, reducing both algorithm complexity and processing time
Solution Approach 2:
Instead of applying refinement steps sequentially after initial haze removal, the patent inverts the approach by performing all necessary calculations in a single pixel-based pass. This inversion eliminates the need for multiple sequential processing stages, reducing algorithm complexity while maintaining image quality
3Reliability
If conventional algorithms use dark channel prior, then haze removal can be performed, but artifacts are introduced and processing time increases
Solution Approach 1:
The patent creates a simplified copy of the essential functionality without the problematic components. By copying the core haze removal function but replacing the dark channel prior with a pixel-based atmospheric light estimation model, the system eliminates blocking artifacts while maintaining haze removal capability through a cleaner mathematical formulation
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method effectively enhances image visibility by reducing computational costs and processing time while avoiding blocked artifacts, ensuring accurate dehazing of hazy images with improved contrast and reduced discoloration.
Implementation Method 1
Light may be scattered and absorbed by these particles causing haze to be present in a captured image
Implementation Method 2
Light may be scattered and absorbed by these particles causing haze to be present in a captured image
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for enhancing an image and an image enhancement device are described. The method for enhancing an image including: capturing an initial image including a plurality of pixels, and performing a pixel-by-pixel dehazing operation for each of the plurality of pixels. The performing including: generating, for each of the plurality of pixels, a value for a blended gray image based on color channels of the pixel, generating, for each of the plurality of pixels, a value for a transmission map based on the blended gray image, and generating, for each of the plurality of pixels, output color channels for a processed image based on the value for the transmission map, the processed image being an enhancement of the initial image.