Research Themes

MIRA focuses on multi-modal spectral imaging and fusion, including calibration, registration, and super-resolution. Below is some description of the different research themes tackled by the platform.

Spectral Imaging

Registration

Super-resolution

Spectral Imaging

We acquire images that go beyond the visible range to capture detailed spectral information for each pixel. Our workflow includes preprocessing and radiometric calibration of raw data to ensure accurate reflectance values and spectral consistency across sensors. This enables reliable analysis of materials, biological tissues, and manufactured products using their unique optical signatures.

Raw HSI — Raw hyperspectral image (uncalibrated)

Calibrated HSI — Radiometrically calibrated reflectance image

Image Registration

We align images from multiple cameras or sensors so that every pixel represents the same point in the scene. This spatial matching is essential for combining data from color, infrared, and hyperspectral cameras accurately, ensuring high-quality fusion and comparison between modalities.

Reference HR image — Reference image (HR)

Target LR image — Target image (LR, before alignment)

Registered LR image — Target image after registration

Super-resolution

We combine information from multiple low-resolution images or sensors to produce sharper, higher-resolution results. This enhances image detail and clarity, enabling improved visualization, quantitative analysis, and advanced applications such as defect detection and material mapping.

Super-resolved image — Super-resolved fusion result

Check here for an example case scenario and interactive demo.

Super-resolution

- Example Case Scenario -

In multimodal imaging, different cameras often capture complementary information. Cam 1 provides rich spectral data but lower spatial detail, while Cam 2 offers sharper spatial resolution with fewer spectral bands. Through super-resolution fusion, these data sources are combined to produce an image that is both spatially detailed and spectrally rich, as summarized below:

Camera / Result	Spectral Resolution	Spatial Resolution	Description
Cam 1	High	Low	Captures detailed spectral information for material analysis.
Cam 2	Low	High	Captures fine spatial structures with fewer spectral bands.
Fused Image	High	High	Combines both to yield enhanced spatial and spectral quality.