Multi-feature Image Haze Removal Using Feature Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dehazing techniques often result in images that appear unrealistic or are not visually pleasing, leading users to avoid haze removal features, which can discourage capturing scenes with haze.
Innovation Solution
The method involves extracting multiple feature maps from hazy images to compute unscattered light and airlight, using a trained model to generate a dehazed image, and applying filters to adjust pixel brightness for a visually pleasing outcome.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional dehazing techniques are used, then haze is removed from the image, but the resulting image appears unrealistic or too dark
Solution Approach 1:
The patent segments the dehazing process into multiple independent feature extraction stages: dark-channel feature extraction, contrast feature extraction, hue-disparity feature extraction, and saturation feature extraction. Each feature map is processed separately to capture different aspects of the hazy image, allowing for more precise control over the dehazing operation and avoiding the unrealistic artifacts produced by conventional single-stage methods.
Solution Approach 2:
The patent transforms the hazy image into multiple feature maps representing different parameters (dark-channel, contrast, hue-disparity, saturation). By operating in this transformed parameter space and then reconstructing the dehazed image from these processed features, the method achieves more realistic results while effectively removing haze, avoiding the overly dark appearance of conventional techniques.
2Object-affected harmful factors
If conventional dehazing techniques are used, then haze is removed from the image, but the resulting image is not visually pleasing
Solution Approach 1:
The patent specifically processes color-related features including hue-disparity information and saturation features. By extracting and processing these color characteristics separately and then reconstructing the image with preserved color information, the method produces visually pleasing results with natural colors, avoiding the unnatural appearance that can result from conventional dehazing.
Solution Approach 2:
The patent adds multiple dimensional representations of the image by creating separate feature maps for different visual characteristics (dark-channel, contrast, hue, saturation). This multi-dimensional approach allows for more comprehensive control over the dehazing process, producing images that are both haze-free and visually appealing by considering multiple aspects of image quality simultaneously.
3Manufacturing precision
If multiple feature maps are extracted and processed, then image realism and visual appeal are improved, but processing complexity increases
Solution Approach 1:
The patent divides the complex dehazing task into separate, manageable feature extraction and processing stages. Each feature map (dark-channel, contrast, hue-disparity, saturation) is extracted and processed independently using dedicated algorithms, making the overall complex process more manageable and allowing for optimized processing of each feature type.
Solution Approach 2:
The patent employs a unified framework that handles multiple feature types through a common processing pipeline. The same basic operations (feature extraction, processing, and image reconstruction) are applied across different feature maps, providing a multi-functional solution that manages complexity through consistency and reusability of processing steps.
Data Source
AI summary
Multi-feature image haze removal is described. In one or more implementations, feature maps are extracted from a hazy image of a scene. The feature maps convey information about visual characteristics of the scene captured in the hazy image. Based on the feature maps, portions of light that are not scattered by the atmosphere and are captured to produce the hazy image are computed. Additionally, airlight of the hazy image is ascertained based on at least one of the feature maps. The calculated airlight represents constant light of the scene. Using the computed portions of light and the ascertained airlight, a dehazed image is generated from the hazy image.