Page 115 - Digital Analysis of Remotely Sensed Imagery

86 Cha pte r T h ree

For instance, storage of a 512 by 512 image of 4 bands that has a
quantization level of 8 bits (256 gray levels) requires a space of
8,388,608 bites or 8.4Mb. This number doubles or quadruples if the
quantization level increases by 1 or 2 bits. Data recorded in 10 bits
and 11 bits that are rather common with the recently emerged
hyperspatial resolution satellite data (see Chap. 2) require more
storage space than other multispectral data.
The above calculation of storage space applies to images whose
pixel values are integers, as with all raw data. However, some
processed results (e.g., band ratioing) are more appropriately saved
in the floating point mode. Storage of such pixel values requires more
space. Instead of a single byte for one pixel value, a pixel value saved
in the float point format requires 4 bytes, thus quadrupling the storage
space calculated using Eq. (3.1).
3.1.2 Data Storage Forms
If saved in the binary format, multispectral satellite images may be
stored in one of three forms, band sequential (BSQ), band interleaved
by line (BIL), and band interleaved by pixel (BIP). In the BSQ form, all
the information related to one particular band and the header
information is stored in one file. Multiple bands are written
sequentially into this file (Fig. 3.1). In case of multispectral bands, the

5 8 2 4 1 1 0 5545999
2 4 3 1 1 0 6 2234446
2 4 4 7 7 1 5 5235565
3 3 4 8 8 9 5 8820033

4 4 5 5 8 8 0 1134009
6 6 2 2 3 3 3 1244778
6 7 7 1 4 5 4 2255734

(a) Band 1 (b) Band 2

5824110243110624477153348
8954455880662233367714545
5459992234446523556588200

33113400912447782255734

FIGURE 3.1 The original satellite imagery of two spectral bands [refer to (a)
and (b) here]. Storage of the two spectral bands in the BSQ format inside the
computer (bottom draw).