AICOOL: A New-Generation Technology for Fully Automatic Haze Removal and Image Clarification

To avoid ambiguity, the term fully automatic atmospheric correction is used here in a strict technical sense:

Definition — Fully Automatic Atmospheric Correction

An atmospheric correction system is considered fully automatic if and only if it meets all of the following criteria:

• No manual parameter tuning: All parameters are determined automatically without user intervention or scene-specific adjustment.
• No human-in-the-loop decision-making: The workflow does not rely on expert judgment, visual inspection, or interactive control during processing.
• Scalable and consistent batch processing: The same correction logic is applied consistently across large and heterogeneous image collections.
• Physics-preserving correction: Radiometric and physical consistency is maintained; results are not produced by purely visual or empirical optimization.
• Automated quality control: Processing quality is evaluated internally using objective criteria, without manual acceptance or rejection.

Only systems that satisfy all of the above conditions qualify as fully automatic atmospheric correction. Methods requiring expert supervision or interactive adjustment do not meet this definition.

AICOOL is a new-generation, fully automatic haze removal and image clarification technology developed for high-resolution optical satellite and aerial imagery. Unlike conventional atmospheric correction methods based on radiative transfer modeling and auxiliary metadata, AICOOL adopts a fundamentally different, image-driven framework to separate atmospheric effects from surface information. At the core of AICOOL is a novel electromagnetic wave component formulation combined with modern mathematical analysis. By exploiting the distinct geometric, structural, and radiometric characteristics of haze, thin clouds, and ground objects, the method is able to decompose complex and spatially heterogeneous atmospheric effects directly from image content — without relying on sensor metadata, calibration parameters, or external atmospheric inputs.

This paradigm enables AICOOL to deliver clear imagery with improved radiometric consistency and spatial fidelity, even under challenging non-uniform haze conditions. The entire processing chain is fully automatic, computationally efficient, and designed for large-scale batch processing, making it well suited for modern Earth observation workflows.

Key Advantages

• New-generation, image-driven approach fundamentally different from radiative-transfer-based atmospheric correction
• Fully automatic processing without sensor metadata or auxiliary atmospheric information
• Robust separation of haze, thin clouds, and surface components under spatially heterogeneous conditions
• Preserves relative spectral and radiometric consistency for quantitative analysis
• High-throughput batch processing capability for large image archives
• Initially developed for Sentinel-2 and scalable to a wide range of high-resolution optical sensors

Technology Foundation

AICOOL is founded on a mathematical decomposition of electromagnetic wave behavior under atmospheric scattering conditions. Rather than estimating atmospheric parameters explicitly, the framework separates atmospheric and surface components by leveraging their intrinsic differences in topology, scale-space behavior, and spatial organization. This fundamentally different formulation allows AICOOL to estimate both haze distribution and surface-related image components simultaneously — a capability that is difficult to achieve using traditional atmospheric correction models. As a result, the method produces clearer imagery and more reliable surface information for downstream analysis.

This fundamentally new framework leads to clearer images, more accurate surface information, and greatly improved usability for downstream machine learning and remote sensing analytics.

Applications

• High-accuracy land cover change detection
• Ground object recognition and monitoring
• Agricultural, forestry, and ecological assessment
• Large-area image mosaicking and radiometric normalization
• Governmental and commercial Earth observation missions requiring reliable, fully automatic preprocessing