Near space hyperspectral interferometric imaging image quality assessment with a physically grounded dataset

Area of Science:

• Atmospheric science and remote sensing.
• Optical imaging and signal processing.

Background:

• Near-space hyperspectral interferometric imaging is vital for atmospheric observation, enabling high-resolution profiling of greenhouse gases and wind fields.
• This imaging modality suffers from nonlinear degradations like angle deviations, vibrations, and sensor non-uniformities, complicating accurate image quality assessment (IQA).
• Existing IQA benchmarks, based on natural images, lack physical realism and domain-specific distortions, leading to poor performance of trained models on interferometric data.

Purpose of the Study:

• To introduce NSIQ, the first image quality assessment benchmark specifically designed for near-space interferometric imaging.
• To provide a realistic and domain-specific dataset for evaluating IQA models in this specialized field.
• To highlight the limitations of current IQA approaches when applied to physics-driven degradations in interferometric systems.

Main Methods:

• Developed NSIQ, a benchmark comprising 201 grayscale interferograms generated via a physics-consistent simulation framework.
• Incorporated six representative degradation types reflecting realistic system-level distortions.
• Annotated each sample with hybrid quality labels, combining expert perceptual scores and normalized physical parameters for a multi-dimensional quality assessment.

Main Results:

• Benchmarking demonstrated that state-of-the-art IQA methods, effective on natural images, exhibit significant performance degradation on the NSIQ dataset.
• The results underscore the inadequacy of current IQA models for the unique challenges posed by near-space interferometric imaging.
• Identified an urgent need for domain-adaptive and physically grounded IQA models tailored to this application.

Conclusions:

• The NSIQ benchmark provides a crucial resource for advancing IQA in near-space interferometric imaging.
• Facilitates research in environmental monitoring, atmospheric modeling, and intelligent remote sensing.
• Establishes a foundation for improved long-term Earth system observation and understanding.