An overview of raster storage settings

Raster storage environment settings can be used to adjust the default compression type, the default settings for pyramid creation and calculating statistics, and the default tile size used by geoprocessing raster tools.

Not all settings apply to all storage types. Refer to the Raster storage matrix (below) for more details. File Group 2 consists of ERDAS IMAGINE files. All remaining ArcGIS supported raster file formats fall into File Group 1.

Raster storage matrix

Storage settings	File Group 1	File Group 2	personal GDB	File GDB	ArcSDE
Pyramids	yes OVR file	yes RRD type	yes RRD type	yes	yes
Resampling	yes	yes	yes	yes	yes
Levels	yes	yes	yes	yes	yes
Raster statistics	yes	yes	yes	yes	yes
skip factor	yes	yes	yes	yes	yes
ignore value	yes	yes	yes	yes	yes
Compression	yes*	yes RLE compression	yes	yes	yes
LZ77	yes*	no	yes	yes	yes
JPEG	yes*	no	yes	yes	yes
JPEG 2000	yes*	no	yes	yes	yes
Tile size	no	no	no	yes	yes

Raster storage matrix

Note:

Compression is dependent on the type of file format. Please refer to the Technical Specifications to see which file formats are able to support compression.

Compression type

The compression type setting is used by any tool whose output is a raster dataset. There are nine different compression methods available for geoprocessing tools. Of these compressions, four types of compression are supported when loading rasters to a geodatabase: LZ77, JPEG, JPEG2000, and NONE.

Valid compressions for each pixel depth

Compression	Pixel Depth (8 bit)	Pixel Depth (16 bit)	Additional information
LZ77	Yes	Yes	Any pixel depth
JPEG	Yes	No
JPEG2000	Yes	Yes
PackBits	Yes	No	1-bit to 8-bit data
LZW	Yes	Yes	Any pixel depth
RLE	Yes	Yes	Any pixel depth
CCITT_G3	No	No	Only for 1-bit data
CCITT_G4	No	No	Only for 1-bit data
CCITT_1D	No	No	Only for 1-bit data

Valid compressions for each pixel depth

LZ77 (the default) is a lossless compression that preserves all raster cell values. It uses the same compression algorithm as the PNG image format and one similar to ZIP compression. As you can rely on the pixels not changing their values after you compress them, use LZ77 for performing visual or algorithmic analysis.

JPEG is a lossy compression, because raster cell values may not be preserved after compression and decompression. It uses the public domain JPEG (JFIF) compression algorithm and only works for unsigned 8-bit raster data (single-band grayscale or three-band raster data).

JPEG2000 uses wavelet technology to compress rasters, so they visually appear lossless, meaning that although the cell values do get manipulated, the differences between the original and the same raster with compression are not easily distinguishable. Use JPEG or JPEG2000 for rasters that are meant as pictures or backdrop imagery.

If JPEG or JPEG2000 is selected, you can also set the compression quality to control how much loss the image will be subjected to by the compression algorithm. The values of the pixels of an image compressed with a higher compression quality will be closer to those of the original image. Valid value ranges of compression quality for JPEG are from 5 to 95. Valid value ranges for JPEG 2000 are from 1 to 100. The default compression quality is 75. The amount of compression will depend on the data and compression quality. The more homogeneous the data, the higher the compression ratio. The lower the compression quality, the higher the compression ratio. Lossy compression normally results in higher compression ratios when compared to lossless compression.

The primary benefits of compressing data are that compressed data requires less storage space and data display times will be quicker, as there is less information to transmit.

Pyramids

Pyramids are reduced-resolution representations of your dataset. They can speed up display of raster datasets by retrieving only the data that is necessary at a specified resolution. By default, pyramids are created for raster datasets by resampling the original data. There are three resampling methods available: nearest neighbor, bilinear, and cubic.

The default is nearest neighbor. It works for any type of raster dataset. Use nearest neighbor for nominal data or raster datasets with color maps, such as land-use data, scanned maps, and pseudocolor images.

Use bilinear interpolation or cubic convolution for continuous data, such as satellite imagery or aerial photography.

If you uncheck Build pyramids, pyramids will not be created with the output raster. Not building pyramids saves storage space but will lead to slower display speeds, especially for larger raster datasets.

If the raster pyramids are build as overviews (OVR), then it is also possible to compress the pyramids with either LZ77 or JPEG. If the pyramids can only be built as a reduced-resolution dataset, then no additional compression options are available.

Statistics

The Statistics option enables you to build statistics for output raster datasets. Statistics are required for your raster dataset to perform certain tasks in ArcMap or ArcCatalog, such as applying a contrast stretch or classifying your data. It is not essential to build statistics if they have not already been calculated, since they are calculated the first time they are needed. However, it is recommended that you calculate statistics for your raster datasets before using them if you want to use certain features that require statistics. The default display of your raster will be improved in most cases if statistics have already been calculated, because a standard deviation stretch is applied if statistics are present.

Setting a Skip factor allows you to speed up the process of calculating statistics by skipping pixels. The Skip factor does not apply for GRID datasets.

Values you set to ignore will not participate in the statistics calculation. Normally, you may want to ignore the values of the background. This only applies to personal geodatabase raster datasets.

Tile size

The tile size setting is used by any tools that create raster datasets and are stored in blocks.

The default tile size is 128 by 128, which is good for most cases. However, if the tile size is too big, you will end up bringing up more data than is needed each time you access the data. For example, you want to display a window of 100 by 100, and it only covers one tile. If you set the tile size to 512, you need to get the tile of 512 by 512 pixels. If your tile size is set to 128 by 128, you'll bring up less extra data if the display window is 100 by 100.