Following are descriptions of some terms, phrases and acronyms used in this documentation that you may not be familiar with.
Background color
The pixel value that defines the color of the image outside the pixel extents of the image. This property is most often used in mosaicking. See also "Nodata color" and "Transparency color".
Band
A set of samples corresponding to one spectral component of an image e.g. the “red” band of an RGB image. Also known as component. There is a weak correlation between the colorspace and the number of bands: grayscale is 1, RGB is 3, CMYK is 4. (Multispectral is the exception, as it has one or more bands.)
BBB
Name of a file format for raw images, short for "BIP/BIL/BSQ."
BIL
One of the three “BBB” formats; abbreviation for "band-interleaved by line." In this data layout, the samples of the image are laid out per line, one sample at a time. Figure 1 shows a 3-banded image of width 3 and height 2 using a BIL layout.
R00 R01 R02 G00 G01 G02 B00 B01 B02 R10 R11 R12 G10 G11 G12 B10 B11 B12 Figure 1: BIL Layout
BIP
One of the three “BBB” formats; abbreviation for "band-interleaved by pixel." In this data layout, the samples of the image are laid out per pixel, one sample at a time. Figure 2 shows a 3-banded image of width 3 and height 2 using a BIP layout.
R00 G00 B00 R01 G01 B01 R02 G02 B02 R10 G10 B10 R11 G11 B11 R12 G12 B12 Figure 2: BIP Layout
BSQ
One of the three “BBB” formats; abbreviation for "band-sequential." In this data layout, the samples of the image are laid out per band, one sample at a time. Figure 3 shows a 3-banded image of width 3 and height 2 using a BSQ layout.
R00 R01 R02 R10 R11 R12 G00 G01 G02 G10 G11 G12 B00 B01 B02 B10 B11 B12 Figure 3: BSQ Layout
Byte order
See "Endianness".
Component
See "Band".
Composite image
With respect to MG3 and MG4, an image made up of several internal images (known as tiles). See also "MG3" and "MG4".
Compression
An operation that creates a new image file from an original image file such that the file size of the new image is smaller. The reduction in file size may be at the expense of some image quality. (Note that the file size is what is reduced, not the dimensions of the image itself.) See also "Lossless" and "Lossy".
Compression ratio
The amount or degree of reduction in file size, expressed as the ratio of the nominal file size to the target size. Because nominal size is used, compression ratios from compressed formats, e.g. JPEG, can be misleading. For example, consider a 50MB raw file compressed to a 10MB JPEG file: this is 5:1 compression ratio. Performing a further 10:1 MrSID compression on that 10MB JPEG file will produce a 5MB MrSID file, not a 1MB file, because the 10:1 is relative to the image's nominal size and not it's current physical file size. See also "Nominal image size".
Delegate
An abstract class used to implement callbacks. The MrSID SDK uses delegates for situations such as detecting interrupts or reporting progress during a decode operation.
Dynamic range
The range values of the samples of an image.
Dynamic range can be expressed in two forms, "min/max" and "window/level". In the first form, the range is expressed using minimum and maximum values. In the second form, the range is expressed with a "window" value equal to the size of the range, for example (max-min+1), and a "level" value equal to the midpoint of the range, for example (min+max)/2. Window and level correspond roughly to the concepts of "contrast" and "brightness", respectively.
Some images, notably those with 16-bit data, often display as "very dark" because the full 16-bit range of the sample size is not used by the actual sample values; dynamic range information can be used to scale or stretch the data for better presentation. The figure below illustrates the dynamic range of a 10-bit image (log scale).
![]()
Figure 4: Dynamic range of a 10-bit image (log scale)
Endianness
the order in which bytes in a multibyte word are stored. In a big endian machine the most significant byte of the word is stored at the lowest address of the word; in a little endian machine, the most significant byte is stored at the highest address. Big endian machines include SPARC and PowerPC. Little endian processors include the Intel x86.
The table below shows the effects of endianness where:
unsigned int value = 0x01020304;
unsigned char* p = (unsigned char*)&value;
Address Little Endian Value Big Endian Value *p 0x04 0x01 *(p+1) 0x03 0x02 *(p+2) 0x02 0x03 *(p+3) 0x01 0x04 Figure 5: Endianness and Byte Order
Frequency balance
A parameter used in MrSID to determine the emphasis given to edges and “flat” color areas when performing compression. In MG2 encoding this parameter is called gamma. A lower value creates more defined edges, while a higher setting creates softer edges. See also "Weight" and "Sharpness".
Gamma
See "Frequency balance".
GML
An acronym for geography markup language, an XML-based language for representing many types of geographic content. Using GML primitives one can describe coordinate reference systems, units of measure, features, geometries and topologies, coverages, annotations and more.
JPEG
A widely used image format that supports image compression. JPEG is an ISO standard. See also JPEG 2000.
JPEG 2000
A new, wavelet-based image format designed for high-quality imagery and advanced imaging workflows. JPEG 2000 is an ISO standard. See also JPEG, MrSID.
K-weight
A parameter used in MrSID to determine the emphasis given to the K (black) band of a CMYK image when performing compression.
Level
See "Zoom level".
Lossless
An image that contains a representation of all of the original pixel values; when decoded, a lossless image is mathematically identical to the original. Note that lossless refers only to the image data and implies nothing about the image metadata. See also "Visually lossless" and "Lossy".
Lossy
An image that contains an approximation of all of the original pixel values; when decoded, depending on the quality of the compression, a lossy image may appear to be a poor representation of the original, visually indistinguishable from the original, or anywhere in between. The perceived image quality is affected by changes in the sharpness of edges, color balance, reduced resolution, and so forth. See also "Lossless" and "Visually lossless".
Magnification
The scale at which an image is represented. Magnification is expressed as a positive floating point value: 1.0 represents the full image (at “one to one”), 0.5 represents a half-scale (lower resolution) version, and 2.0 represents a double-scale (expanded) resolution. The magnification value must be a power of two. See also "Resolution", "Scale" and "Zoom level".
Metadata
Information about the image, as opposed to the actual image data (pixels). Metadata can refer to basic image properties such as width, height, colorspace, etc., but generally refers to less fundamental properties such as geospatial position, name of image creator, image date, etc.
MG2
MrSID Generation 2. The earliest released version of the MrSID image format. MG2 is limited to lossy encoding and does not support optimization.
See also "MrSID", "MG3" and "MG4".
MG3
Mrsid Generation 3. The second released version of the MrSID image format. MG3 supports lossless encoding, image optimization, and composite images. See also "MrSID", MG2 and "MG4".
MG4
Mrsid Generation 4. The current version of the MrSID image format. MG4 supports most of the features that MG2 and MG3 support, but also supports alpha channels, multispectral and hyperspectral imagery and improved composite mosaics. See also "MrSID", MG2 and "MG3".
Mosaic
A composition of two or more images to form a new, larger image. Positioning of the images is generally based on geospatial coordinates.
MrSID
Acronym abbreviating Multiresolution Seamless Image Database. MrSID is a wavelet-based image format designed for large, high-quality geospatial imagery. There are two versions of MrSID, known as MG2, MG3 and MG4. See also "MG2", "MG3" and "MG4".
Nodata color
Same as "Transparency color". The pixel value that defines the color of the image that corresponds to a region of the image that has no valid data. Note that a transparency region of the image is contained with the image extents, whereas the background color is used for regions outside of the image. This property is most often used in mosaicking. See also "Background color".
Nominal image size
The size of an image, in bytes, defined by the product of
image width x image height x number of bands x number of bytes per sample
See also "Physical image size".
Optimization
The process of creating an MG3 or MG4 image from a source MG3 or MG4 image, such that the new image is better suited for some purpose or workflow. The most common optimization is compression; other optimization operations include cropping and removal of resolution levels.
Physical image size
The size of an image, in bytes, as defined by the file size required to represent the image on disk. See also "Nominal image size".
Pixel
A set of samples, together making up a single (x,y) point in the image; for example, a 100 x 200 RGB image will contain 20,000 pixels. See also "Sample".
Resolution
Definition #1) The scale at which an image is represented as expressed in a wavelet level. The pixel resolution at level 0 (zero) is equal to a magnification of 1.0. See also "Magnification", "Zoom level" and "Scale".
Definition #2) Ground units per pixel as used for georeferencing (a value stored in the metadata of an image).
Sample
A value representing a magnitude or intensity, for example a red sample in an RGB pixel. A 100 x 200 RGB image will contain 60,000 samples. See also "Pixel".
Scale
The resolution or magnification at which an image is represented. Scale is represented as a signed integer, corresponding to the negative of the log of the magnification. That is, magnifications of 1.0, 0.5, and 2.0 are equivalent to scales of 0, 1, and -1, respectively. See also "Magnification", "Zoom level" and "Resolution".
Scene
The region of an image to be decoded, as defined by an upper-left position, width and height dimensions, and magnification.
Sharpness
A parameter used in MrSID to determine the sharpness of boundaries between different areas of an image when performing compression. See also "Frequency balance" and "Weight".
Stream
An abstract interface to an arbitrary chunk of data, represented as an array of bytes; examples of data that can be represented as streams include files, memory buffers, and sockets.
Strip height
The number of rows of an image to be processed in each step of an image read operation. Use of smaller strip heights may reduce memory requirements, but at a possible performance loss.
Transparency color
Same as "Nodata color". See also "Background color".
Visually lossless
An image that contains a close approximation of all of the original pixel values; when decoded, a visually lossless image appears to a typical viewer as being indistinguishable from the original. A visually lossless image is a lossy image. See also "Lossless" and "Lossy".
Wavelet
A mathematical representation of a pixel value that varies by frequency and duration; in wavelet space, the importance of a pixel depends on the values of its neighboring pixels. Wavelet-based image formats are able to present images at multiple levels of resolution without the overhead of pyramidal formats.
Weight
A parameter used in MrSID to determine the emphasis given to the grayscale portion of a color image when performing compression. See also "Frequency balance", "Sharpness" and "K-weight".
World file
A text file that contains geospatial positioning information to augment or replace the geospatial information in an image file.
Zoom level
The scale at which an image is represented. Levels are generally expressed with signed integer values. An image “at scale 1” has half the width and height of the original. See also "Scale", "Magnification" and "Resolution".