Haze-Aware Single-Image Clarity Enhancement
Official patent title
Apparatus and method for single image dehazing considering haze type classification
Arabic title: جهاز وطريقة لإزالة الضباب من صورة واحدة مع مراعاة تصنيف نوع الضباب
Invention
Invention
Problem
A single dehazing model may handle fog, smoke, cloud, rain, and environmental haze unevenly because each condition alters contrast, color, and detail differently.
Why it matters
Better visibility restoration can improve the input quality available to downstream computer-vision tasks such as detection, segmentation, remote sensing, and surveillance analysis.
Approach
The system classifies the haze type with a selected single classifier or a hybrid conditional classifier, chooses a dehazer specialized for the predicted class, and applies it to the input image.
Who may benefit
Potential beneficiaries include computer-vision developers, autonomous-system teams, surveillance and remote-sensing providers, agricultural-imaging groups, and image-processing researchers.
Potential value
The workflow separates haze recognition from haze removal, then routes each image to a specialized model, allowing the classifier set, haze classes, intensity levels, and application domain to be adjusted.
Background
Background
Atmospheric haze from fog, clouds, pollution, smoke, rain, or other particles reduces image contrast, shifts color, and hides detail. This can degrade not only visual quality but also automated navigation, surveillance, remote-sensing, and other computer-vision pipelines. Existing datasets and dehazing approaches often focus on one haze type or a narrow set of intensity levels, which limits classification and targeted restoration across changing conditions. The patent creates diverse synthetic training sets and uses haze-type prediction to select a corresponding specialized dehazer.
Technology overview
Technology overview
Ground-truth images are divided into outdoor, street, farmland, and satellite sets and overlaid with synthetic haze at selected intensity levels. The training data cover cloud, fog, environmental haze, rain, and, for selected sets, smoke. Multiple classifiers and specialized dehazers are trained. The single selective classifier uses one chosen model, while the hybrid conditional classifier combines predicted classes, probabilities, class count N, and model count M through condition blocks to select a final haze class. That class determines which specialized dehazer processes the image.
Potential applications
Potential applications
- Potential use in camera pipelines for autonomous vehicles or driver-assistance research.
- Potential use in surveillance-camera image enhancement.
- Potential use in satellite, aerial, and remote-sensing imagery.
- Potential use in farmland and outdoor computer-vision systems.
Evidence-supported advantages
Evidence-supported advantages
- Classifies haze type before selecting a specialized restoration model.
- Covers cloud, fog, environmental haze, smoke, and rain conditions.
- Offers single-model selection and multi-model hybrid conditional classification.
- Allows datasets, model counts, haze classes, and intensity levels to be adjusted.
Development stage
Development stage
Patent publication describing dataset generation, model training, and an inference workflow; independent benchmark performance, deployment, and commercialization were not established.
Commercial opportunity
Commercial opportunity
The architecture may support licensing or integration with camera, autonomous-system, surveillance, or remote-sensing software. Evaluation should use representative real-world data and measure classification error, restoration quality, detail preservation, latency, compute and memory demand, robustness to unseen weather, downstream-task impact, model maintenance, cybersecurity, and privacy obligations.
Patent classifications
Patent classifications
WIPO IPC
- G06T5/60Image data processing or generation, in general
- G06V10/94Image or video recognition or understanding
CPC
- G06V10/945Image or video recognition or understanding
- G06T5/73Image data processing or generation, in general
- G06T5/60Image data processing or generation, in general
- G06V20/50Image or video recognition or understanding
- G06V10/764Image or video recognition or understanding
- G06V10/87Image or video recognition or understanding
- G06V10/774Image or video recognition or understanding
- G06T2207/20081Image data processing or generation, in general
- G06T2207/30232Image data processing or generation, in general
- G06T2207/10032Image data processing or generation, in general
- G06T2207/20092Image data processing or generation, in general
- G06T2207/20084Image data processing or generation, in general
- G06T2207/30181Image data processing or generation, in general
- G06T2207/30252Image data processing or generation, in general
Inventors
Inventors
- First inventorMd Tanvir Islam
- InventorIk Hyun Lee
- InventorAbdul Khader Jilani Saudagar
- InventorAbdullah Altameem
- InventorMohammed Abaoud
- InventorKhan Muhammad
Keywords
Keywords
- single-image dehazing
- haze classification
- computer vision
- hybrid conditional classifier
- specialized dehazer
- synthetic haze
- remote sensing
- surveillance
Patent document and drawings
Patent document and drawings
The patent publication is mapped to this record. Patent drawings remain within that publication; no separately cleared public media package has been supplied.