Quiro libros

Remote Sens. 2010, 2, 1644-1661; doi:10.3390/rs2071644
OPEN ACCESS

Remote Sensing
ISSN 2072-4292 www.mdpi.com/journal/remotesensing Article

Normalizing and Converting Image DC Data Using Scatter Plot Matching
Stephan J. Maas 1,* and Nithya Rajan 2
1

2

Department of Plant and Soil Science, Texas Tech University, and Texas AgriLife Research, 3810 4th Street, Lubbock, TX 79405, USATexas AgriLife Research and Extension Center, 11708 Highway 70 South, Vernon, TX 76384, USA; E-Mail: nithya.rajan@ttu.edu

* Author to whom correspondence should be addressed; E-Mail: stephen.maas@ttu.edu; Tel.: +1-806-723-5235; Fax: +1-806-723-5272. Received: 20 April 2010; in revised form: 11 May 2010 / Accepted: 7 June 2010 / Published: 24 June 2010

Abstract: Remote sensing image data fromsources such as Landsat or airborne multispectral digital cameras are typically in the form of digital count (DC) values. To compare images acquired by the same sensor system on different dates, or images acquired by different sensor systems, it is necessary to correct for differences in the DC values due to sensor characteristics (gain and offset), illumination of the surface (a function of sunangle), and atmospheric clarity. A method is described for normalizing one image to another, or converting image DC values to surface reflectance. This method is based on the identification of pseudo-invariant features (bare soil line and full canopy point) in the scatter plot of red and near-infrared image pixel values. The method, called “scatter plot matching” (SPM), is demonstrated bynormalizing a Landsat-7 ETM+ image to a Landsat-5 TM image, and by converting the pixel DC values in a Landsat-5 TM image to values of surface reflectance. While SPM has some limitations, it represents a simple, straight-forward method for calibrating remote sensing image data. Keywords: remote sensing; surface reflectance; ground cover; soil line; calibration; pseudo-invariant features; canopyreflectance; Landsat

Remote Sens. 2010, 2 1. Introduction

1645

The image data produced by remote sensing systems such as Landsat TM or airborne digital cameras are typically in the form of digital count (DC) values. DC values are related to the reflectance of the observed surface but are affected by other factors, including sensor characteristics (gain and offset), illumination of the surface (afunction of sun angle), and atmospheric clarity. To compare images acquired by the same sensor system on different dates, or images acquired by different sensor systems, it is necessary to correct for differences in the DC values present in the images as a result of differences in the previously mentioned factors. This can be accomplished either by absolute radiometric correction or relativecalibration [1]. In absolute radiometric correction, DC values are converted to values of surface reflectance. In relative calibration, the DC values of one image are normalized to those of another image to provide a relative correction for sensor, illumination, and atmospheric differences. For satellite imagery, correction of DC values for sensor characteristics and sun angle is relativelystraight-forward by calculating corresponding values of exoatmospheric reflectance based on known sensor gain and offset parameters and the time of day and day of year [2]. A similar procedure can be used to calculate at-sensor reflectance for aircraft imagery. The remaining problem is to correct these values for the effects of the atmosphere along the observing path. Numerous approaches for atmosphericcorrection of satellite and aircraft imagery have been described and compared [1,3-7]. These approaches can be separated into two general groups: physically based correction methods, and image-based correction methods. Physically based correction methods use some form of radiative transfer model (RTM) to explicitly compute the effects of atmospheric scattering and absorption on the electromagnetic...