Dark Channel Dehazing for Stable ADAS Camera Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced driver assistance systems (ADAS) and vehicle cameras struggle to effectively process images in poor environments due to issues like brightness saturation and locally degraded brightness, leading to instability in providing accurate road condition information.
Innovation Solution
An image processing method that determines a source transmission map based on a dark channel map, applies different filters to generate transformed transmission maps, and blends haze-free images to produce a stable output image, using a multi-directional kernel-based filter for texture restoration and a guided filter to suppress halo artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a typical ADAS and vehicle camera are used in poor environments, then the system can operate, but the image quality degrades with brightness saturation and locally degraded brightness
Solution Approach 1:
The patent segments the transmission map processing into multiple filters (first filter for texture restoration, second filter for halo artifact suppression). Each filter processes the transmission map to generate a corresponding haze-free image, and these segmented processing results are then blended to produce the final output image. This segmentation allows specialized handling of different image degradation aspects.
Solution Approach 2:
The patent applies different filters to different aspects of the transmission map: a first filter (multi-directional kernel-based filter) specifically for texture restoration and a second filter (guided filter) specifically for halo artifact suppression. This local quality approach ensures that each region of the image receives appropriate processing tailored to its specific degradation characteristics.
2Manufacturing precision
If multiple filters are applied to generate transformed transmission maps, then texture restoration and artifact suppression improve, but the processing complexity increases
Solution Approach 1:
The patent merges the results from multiple filters through a blending process. The first haze-free image (from texture restoration filter) and the second haze-free image (from halo artifact suppression filter) are blended together to generate the final output image. This merging combines the benefits of multiple specialized filters while managing complexity through a unified blending operation.
Solution Approach 2:
The blending operation serves multiple functions: it combines the outputs of different filters, manages the transitions between processed regions, and produces the final haze-free image. This multi-functionality reduces the need for separate processing stages for each filter output.
3Ease of manufacture
If haze removal is performed based on a single transmission map, then the processing is simple, but artifacts like blocking and edge degradation occur
Solution Approach 1:
The patent segments the haze removal process into multiple transmission map transformations, where each transformed transmission map is processed by a different filter. This segmentation prevents the formation of artifacts by distributing the processing load across multiple specialized filters rather than relying on a single transmission map.
Solution Approach 2:
The patent creates a composite processing approach by combining multiple transformed transmission maps (first transformed transmission map, second transformed transmission map) into a blended output. This composite method leverages the strengths of different filter types to produce a final image that avoids the artifacts associated with single-map processing.
Data Source
AI summary
An image processing method includes: determining a source transmission map based on a dark channel map of an input image; determining transformed transmission maps by applying different filters to the determined source transmission map; generating haze-free images by removing haze from the input image based respectively on the determined transformed transmission maps; and generating an output image by blending the generated haze-free images.