Product Processing

Pixxel provides satellite imagery in the following product bundle:

• L2A : Map-projected Bottom-of-Atmosphere (BOA) reflectance image (This is the default product unless otherwise specified).

Name	Description	Product Level
Bottom-of-Atmosphere (BOA) Reflectance	Radiometric, geometric, and atmospheric (aerosol and water vapor) corrected data. The imagery is orthorectified and projected in WGS84. Delivered in GeoTIFF format with accompanying metadata.Pixel values are scaled between 0–50000. To convert values to physical reflectance (0–1), divide all pixel values by 50000.	L2A

Image Processing and Correction

The image processing workflow consists of four major steps: radiometric correction, geometric correction, atmospheric correction, and quality analysis. Before radiometric correction, the raw data downlinked from the satellite is converted to an L0 level for long-term storage. The L0 level data is then used for radiometric correction, where DNs are converted to radiance, and basic striping and high-frequency noise are removed from the data. This data is called L1A. Post L1A, a geometric correction model is applied to georeference and orthorectify the data, storing it as a level L1B. L1B data is used by Pixxel’s internal atmospheric correction model and is converted to TOA reflectance (L1C) and BOA reflectance (L2A) data sequentially. Upon completion, a QA/QC check is performed on each image, observations are recorded, and data is stored for delivery.

Isofit Model

Isofit is the model used to process Pixxel’s hyperspectral imagery. It performs atmospheric and sensor correction to transform raw satellite data into accurate surface reflectance, enabling reliable analysis across various applications.

By modeling the interaction of light as it passes through the atmosphere and is captured by the sensor, Isofit ensures that the resulting imagery accurately represents real ground conditions—crucial for use cases in agriculture, environmental monitoring, and more.

Isofit is highly flexible:

• It can operate on both individual pixels and entire images
• It supports multiple atmospheric models and sensor types, making it adaptable to diverse datasets and conditions

Processing Steps

Step	Description
Raw data	This is raw data downlinked from the satellite.
L0 data	This level is for the long-term archiving of raw data with additional information added at a ground station.
NUC correction	Non-uniformity coefficients (NUCs) are derived during pre-launch calibration and applied to remove non-uniformities from the image. This step includes gain and dark offset corrections, which flatten the detector response and normalize all pixels to a zero baseline, ensuring uniform radiometric performance across the image.
Striping and high-frequency noise correction	Minor striping noise and high-frequency noise present in the data post-NUC implementation are corrected through a filtering approach.
Geo-referencing and Orthorectification	The radiometrically corrected data is geo-referenced and orthorectification is performed by using the Copernicus DEM/terrain models. The output image of this step is referred to as level L1B.
TOA radiance to TOA reflectance	The georeferenced and orthorectified TOA radiance image is converted to a TOA reflectance image using extraterrestrial solar irradiance, Earth-Sun distance, and solar zenith angle. This TOA reflectance data level is termed L1C.
Aerosol correction	The L1B and L1C data are used to estimate and correct the effects of aerosol and water vapor. This step removes the haze effect from the image, making it visibly clearer.
QA/QC	The L2A, L1C, and L1B data undergo automated and random manual quality assessments, and spectral and geometric quality observations are recorded.