Pre-Processing of Optical Imagery

Keywords

• spectrometer
• pre-processing
• calibration
• radiometric
• atmospheric
• geometric
• SNR
• NER
• noise.

Introduction

The pre-processing chain from raw image data to a fully georeferenced and parametrized (calibrated and in units that have a physical meaning) data that an end user typically would deploy for environmental studies is the scope of this chapter. Many calibration and processing steps are similar for multi-band, multispectral and high-spectral resolution imagery. Hence, the processing chain of hyperspectral remote sensing imagery is chosen as a proxy for the broader range of optical, visible, and infrared, data sets. Active sensor systems are excluded in this review as these require a significantly different pre-processing. Thermal infrared imagery is, however, included as it shows similar characteristics to data acquired in the visible and shortwave infrared regions, despite the fact that the physics and engineering behind this type of imagery is different.

The objective of hyperspectral remote sensing (e.g., imaging spectrometry/spectroscopy) is typically to measure quantitatively the components of the Earth system, such as radiance, upwelling radiance, emissivity, temperature, and reflectance, from calibrated spectra acquired as images for scientific research and applications (see also Schaepman, in this volume). These measurements facilitate the estimation of biophysical parameters such as carbon balance, yield/volume, nitrogen, cellulose, chlorophyll, soil constituents, inherent optical properties of water and water quality, optical properties, and atmospheric absorption characteristics of ozone, oxygen, water vapor, and other trace gases.

In the following sections, the pre-processing chain (based on van der Meer and De Jong 2001, and De Jong and van der Meer 2004) and individual pre-processing steps are outlined. Radiometric calibration, geometric aspects, and atmospheric influences will be discussed concluding with an evaluation of inherent sensor and image noise and image defects. Table 16.1 provides an overview of the full processing chain. The intermediate steps highlighted in the table are discussed below, in the same order.

The Pre-Processing Chain

Introduction

Pre-processing operations are intended to correct for sensor- and platform-specific radiometric and geometric distortions of data. Radiometric corrections may be necessary due to variations in image illumination and viewing geometry, atmospheric conditions, and sensor noise and [Page 230]response. Each of these may vary depending on sensor and platform characteristics and conditions during data acquisition. For comparative purposes and for quantification of environmental variables, it is desirable to convert data to known (absolute) radiance, L, in W m-2 sr-1, which can be defined as dΘ = L dA dΩ cosθ for an area dA, arriving at a direction θ to the normal of dA in the range of directions forming a solid angle of dΩ steradians or scaled reflectance units. Radiance can be integrated over a specific range of wavelengths, spectral radiance, and expressed as the photon flux power per unit area per solid angle per unit of wavelength interval. Martonchik et al. (2000) and Schaepman-Strub et al. (in this volume) provide reviews of nomenclature of units used in remote sensing. From an engineering perspective, quantifying radiance requires knowledge of the following sensor characteristics:

Table 16.1 Overview of Pre-processing Chain for Spectrometer Data (after Peter Strobl Personal Communication)

...