52
2: Raghuveer M. Rao, Manoj K. Arora
2.2
Image File Formats
Digital images are organized into records. Appropriate header data indicate
the organization. The earliest storage medium for remotely sensed images
was a magnetic tape. Consequently, several image format schemes are tailored
towards the sequential access mode of this medium. With the increased use of
optical storage media, some of the schemes have fallen out of favor. Because
hyperspectral data are gathered in terms of spatial location, in the form of
pixel coordinates, and spectral bands (i. e. gray values of the pixel in different
bands), image format schemes typically differ in how they handle data with
respect to location and bands.
The band interleaved by pixel format (BIP) is pixel oriented in the sense that
all data corresponding to a pixel are first stored before moving on to the next
pixel. Thus, the first record corresponds to the first pixel of the first band of
the first scan line. The second record corresponds to the first pixel of the first
line but data in band 2 and so on. After data from all the bands are recorded
sequentially, we move on to pixel 2 of the first line.
In the band interleaved by line format (BIL) the data are organized in lineoriented fashion. The first record is identical to that of the BIP image. However,
the second record corresponds to the second pixel of band 1 of the first line.
Once all pixel values of band 1 ofline 1 have been recorded, we begin recording
data of pixel 1 of band 2 followed by pixel 3 of band 2 and so on. We begin
recording of line 2 data after completion of recording of all line 1 data.
The band sequential format (BSQ) requires storing of the data of each band
as a separate file. This format is most useful when one might process just one
or a few bands of data. With the availability of random access capability in
optical storage media, such a separation into band-based image files does not
provide a substantial increase in search speed over the other formats when one
is interested in a small number of bands.
A versatile data format called the hierarchical data format (HDF) has been
developed for the National Center for Supercomputing Applications with the
aim of facilitating storage and exchange of scientific data. This format provides
for different data models and supports single file and multifile modes. The following different data structures are supported in HDF4: raster images, color
palettes, texts, scientific data sets (which are multidimensional arrays), Vdata
(2D arrays of different data types) and Vgroup (a structure for associating
sets of data objects). In the newer HDF5, only two fundamental structures are
supported: groups and dataspaces. HDF-EOS an extension ofHDF5 developed
by NASA for the Earth Observing Systems, additional structures are used.
These are point, swath and grid. Some advantages of HDF are: portability to
platform independence, self-documentation, efficient storage and access, accommodation of multiple data types within the same file and ease of extending
the format to suit specific needs (Folk et al. 1999; Ullman 1999).
2: Raghuveer M. Rao, Manoj K. Arora
2.2
Image File Formats
Digital images are organized into records. Appropriate header data indicate
the organization. The earliest storage medium for remotely sensed images
was a magnetic tape. Consequently, several image format schemes are tailored
towards the sequential access mode of this medium. With the increased use of
optical storage media, some of the schemes have fallen out of favor. Because
hyperspectral data are gathered in terms of spatial location, in the form of
pixel coordinates, and spectral bands (i. e. gray values of the pixel in different
bands), image format schemes typically differ in how they handle data with
respect to location and bands.
The band interleaved by pixel format (BIP) is pixel oriented in the sense that
all data corresponding to a pixel are first stored before moving on to the next
pixel. Thus, the first record corresponds to the first pixel of the first band of
the first scan line. The second record corresponds to the first pixel of the first
line but data in band 2 and so on. After data from all the bands are recorded
sequentially, we move on to pixel 2 of the first line.
In the band interleaved by line format (BIL) the data are organized in lineoriented fashion. The first record is identical to that of the BIP image. However,
the second record corresponds to the second pixel of band 1 of the first line.
Once all pixel values of band 1 ofline 1 have been recorded, we begin recording
data of pixel 1 of band 2 followed by pixel 3 of band 2 and so on. We begin
recording of line 2 data after completion of recording of all line 1 data.
The band sequential format (BSQ) requires storing of the data of each band
as a separate file. This format is most useful when one might process just one
or a few bands of data. With the availability of random access capability in
optical storage media, such a separation into band-based image files does not
provide a substantial increase in search speed over the other formats when one
is interested in a small number of bands.
A versatile data format called the hierarchical data format (HDF) has been
developed for the National Center for Supercomputing Applications with the
aim of facilitating storage and exchange of scientific data. This format provides
for different data models and supports single file and multifile modes. The following different data structures are supported in HDF4: raster images, color
palettes, texts, scientific data sets (which are multidimensional arrays), Vdata
(2D arrays of different data types) and Vgroup (a structure for associating
sets of data objects). In the newer HDF5, only two fundamental structures are
supported: groups and dataspaces. HDF-EOS an extension ofHDF5 developed
by NASA for the Earth Observing Systems, additional structures are used.
These are point, swath and grid. Some advantages of HDF are: portability to
platform independence, self-documentation, efficient storage and access, accommodation of multiple data types within the same file and ease of extending
the format to suit specific needs (Folk et al. 1999; Ullman 1999).