This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: radiometric corrections, image enhancements, image filtering, data reduction techniques and image classifications. This chapter focuses mainly on processing of remote sensing images acquired in the optical wavelength such as Landsat TM, Landsat MSS and SPOT because environmentalists will mainly make use of these types of images. 5.2 Digital Image Structure A digital recorded image consists virtually of a number of layers (or matrices) of numbers or digital numbers. The intensity value of a pixel (reflectance or thermal emission) is recorded as digital number or DN. Each layer represents a spectral band and each layer consists of a number of columns and rows (i.e. image lines). Each number per layer is the Digital Number (the measured radiance) per picture element or pixel. The measured radiance is mostly stored as a byte (8 bits), values can vary from 0 to 255 and calibration data are needed to transfer them to reflectance or absolute radiance. Sometimes, digital recorded data are stored as 10 bits e.g. NOAA-AVHRR or 16 bits (words) e.g. AVIRIS. Landsat data are stored as a byte. A Landsat TM image consist of 7 layers or matrices (because it has 7 spectral channels) and it has approximately 6165 image lines with approximately 6165 pixels on each image line. Figure 5.1 and 5.2 illustrates the structure of an Landsat TM scene. The structure of a digital recorded image is sometimes called an ‘image cube’. The X- and Y- coordinates within the matrix of DN represent the position of each pixel in the field. The Z- 63 coordinate represents the spectral position (1 to 7 for TM). As is already discussed, the spatial resolution of the thermal band of Landsat TM is 120 by 120 m. When it is stored in a digital form e.g. on Computer Compatible Tape (CCT) it is resampled to a pixel size of 30 by 30 m. Figure 5.1 Arrangement of scan lines and pixels in Landsat images (Sabins, 1987). Line dropouts As was already discussed in chapter 1, Landsat MSS and Landsat TM register radiance 6 lines at a time respectively 16 lines at a time. On some image data of one image line are missing due to (temporal) failure of one of the detectors. As a result every 6th or 16th line appear black. These lines are called periodic line dropouts. The data in these scan lines is lost. Most image processing systems try to correct these lines by averaging the line above and below the lost line. Striping Striping in an image refers to a regular pattern of (horizontal) stripes in an image. Most often these stripes are caused by a slight difference in calibration of the 6 (for MSS) or 16 (for TM) sensors. A slight difference in sensitivity of the sensors might cause ‘striping patterns’ in the image. The effect can play an important role while sensing dark surfaces such as water bodies....
View Full Document
- Winter '12
- Remote Sensing, Electromagnetic spectrum, µm, Infrared