Mosaic datasets consist of the boundary, footprint, and image layers. The boundary layer shows the extent of all the raster datasets in the mosaic dataset as a single polygon, which is multipart if the collection of imagery is not contiguous. The footprint layer shows the extent of each item in the mosaic dataset as a distinct polygon. The footprint attribute table is the catalog of all the images in the mosaic dataset in addition to any associated overviews. In this table, you can sort the imagery based on any of the attributes, such as cloud cover or acquisition date, or any of the sensor characteristics. The image layer controls the rendering of the mosaicked image similar to a raster layer. Display and rendering properties—such as stretch, band combination, resampling, and mosaic method—are modified on this layer.
Select layers from a collection of imagery
The contextual Data tab displays options for selecting the specific layers with which you want to work. If you have a collection of images over a city that comes from a variety of sensors taken at different times of the day on different dates, for example, you can make a selection based on any of these characteristics. This is also known as querying by attribute. In the southwestern United States, much of the annual rainfall occurs over a few weeks during the monsoon season, and you created a mosaic dataset
of Landsat imagery over New Mexico. In this collection, you have multiple scenes for each date that cover the entire state, and those images were collected on different dates. You may want to see images taken during a certain season, or you may want to look at scenes that have minimal cloud cover. You can make these types of selections using Select By Attributes 
.
In addition to making a selection based on an attribute, you can make selections by drawing around the scenes in question. The Rectangle tool 
is good for imagery since typically you don't work with scenes with irregular shapes. Once you've made the selection, you have a choice of how you want to display those scenes. For example, you selected all of the Landsat scenes with minimal cloud cover from the months of July through August and you want to compare what the vegetation looks like for the rest of the year. After making the first selection, you can group the selected imagery and add it to the map. You can then switch the selection, which selects all of the images from January through June and September through December, and add this to the map as a different group.
Use the Raster Item Explorer pane
The Raster Item Explorer option 
can be used to filter items from a large collection of imagery in either a mosaic dataset or an image service and explore the properties of individual items, add them to a map or scene, and view and edit the processing applied to an item.
To learn more, seeRaster Item Explorer.
Display overlapping imagery
MODIS collects imagery systematically in which the imagery is collected according to a defined periodicity and grid alignment; however, with a resolution of 250 meters, those pixels may be too coarse for some remote sensing questions. Conversely, when working with higher-resolution imagery, the revisit cycle is not systematic, and you may get imagery that overlaps in some areas. Deciding how to resolve conflicts from overlapping imagery is a component of working with collections of imagery. You can set up rules that apply to each layer as a whole, or you can apply rules to only the parts of the imagery that are overlapping. These rules are known as sort methods and are located on the Sort drop-down list.
Sort using mosaic methods
Some of the sort methods are well suited to have a collection of imagery be just a collection of imagery. You don't need to perform analysis; you only need to have the imagery in a logical order. In this case, the North-West and None methods are good options. Whenever there is overlap, the North-West method 
resolves it by selecting the layer closest to the northwest corner of the boundary. The None method 
orders the images in the same order as the attribute table.
Other sort methods can be used for zooming in and panning around a collection of imagery. For example, when you choose the Closest To Center method 
, you will be looking at the images closest to the center of the screen. As you zoom in to a feature, the selection updates automatically, so you see the imagery closest to the feature. The Seamline method 
is used for moving around the imagery. When you put a collection of imagery together, there are a couple of options for identifying seamlines. They can be based on the footprints of the imagery or the features within the imagery. However it is set up, when you choose this option, it generates the seamlines, ranks the images based on the distance to those seamlines, and selects the closet imagery. The result is a smooth transition from one image to the next.
High-resolution imagery may be taken from a viewing angle that is not directly overhead. There are advantages to this and one is that it allows you to measure the heights of buildings based on their shadows, creating a more natural-looking image. You can also see the sides of buildings rather than only the roofs. The drawback is that all of the angles are different, which can be disorienting as you pan from one image to another. Choose the Closest To Nadir method 
to select the imagery closest to being viewed from directly overhead. Assuming that all of the imagery is oblique, you end up with a less abrupt transition from one scene to the next if you have a lot of tall features, because the viewing angles are fairly close together.
Choose the By Attribute method 
to use specific attributes to select imagery with minimal cloud cover or that was taken near a certain time. You can then use the Lock Raster method 
to display only the images you select. These methods have the same effect as using the selection tools of the same names.
If you have imagery over an area using multiple flight paths, you can use the Closest To Viewpoint method 
. These paths are stored in the metadata, and they allow you to view the imagery from a number of different angles, so you can see the imagery from all sides.
Resolve overlapping pixels
After you've determined the method for ordering, you can fine-tune the results. Use the Resolve Overlap button 
to access a set of mosaic operators. The First operator ensures that the images at the top of the list is drawn first. Its counterpart is Last, which starts at the bottom of the same list. If you choose Closest To Nadir , all the layers are ranked based on their viewing angle and the layer with a value closest to 0 degrees is returned.
You can also work at the pixel level. There are different averaging and statistical approaches available. The Minimum and Maximum operators take the lowest and highest valued pixel, respectively, of all the overlapping layers. Using this option means that you have no guarantee of looking at the pixels of just one image in the overlapping area but rather a combination of potential layers.
Other options are Blend and Mean, which average the values of the pixels in question. While Mean is the standard averaging technique, Blend works by giving more weight to pixels closer to neighboring images to produce a smoother image. Using either of these operators means that you may not be looking at pixel values from any of the layers.
The last option is the Sum operator, which adds the values of all the overlapping pixels. If you assume that most pixel values are relatively close to one another for each layer, this technique can be useful to visualize where the most layers are.
Compress the imagery
When you share imagery, the amount of data you transmit matters. These collections can overwhelm the download capabilities of those who are accessing your imagery. ArcGIS offers the following options for compressing data:
- None—The data will not be compressed. The advantage of this option is that there is no data loss and it represents the highest possible quality. The drawback is that you will be transmitting the data in the slowest possible form. If you know your users have a reliable network connection, speed may not be an issue.
- JPEG—A compression ratio of 3:1 to 8:1, with little degradation of the image quality, is used. When using this option, you can use the Quality setting to specify a value from 0 to 100. A value of 80 tends to retain image quality while providing approximately eight times the compression.
- LZ77—A lossless compression method that is ideal for scanned maps, classified imagery, and discrete data is used.
- LERC—This option is recommended for data with a large pixel depth, such as float, 32-bit, 16-bit, and 12-bit data. With this option, the quality value represents the maximum error value that is applicable per pixel (not an average for the image). This value is specified in the units of the mosaic dataset. For example, if the error is 10 centimeters and the mosaic dataset is in meters, enter 0.1. LERC compresses better (5 to 10 times) and faster (5 to 10 times) than LZ77 for float data and is better with integer data. With integer data and an error limit of 0.99 or less, LERC behaves as a lossless compression method.
Visualize the information layers
Processing templates are used to generate on-the-fly information layers. When you create a function chain, it's often built from a specific dataset, but you can also use a raster variable in place of that dataset so it can be applied to similar datasets. When you use that variable, the function chain is now referred to as a processing template. Vegetation indexes, elevation products (such as slope, aspect, and hillshade), and image classification results examples of the kinds of information layers you can derive using raster functions.
You can add processing templates to a mosaic dataset or publish an image service with a processing template, allowing you to display the imagery as information layers. When you add a mosaic dataset to a map or connect to an image service, you have access to these information layers using the Processing Templates drop-down menu 
. When you select one of the items, it immediately renders on the display. You can also do the same from the Processing Templates tab on the layer properties page.
Additionally, you can use a processing template to process a mosaic dataset or image service using the Add Custom option on the Processing Templates drop-down menu or the Browse button on the Processing Templates tab on the layer properties page. Both allow you to choose a processing template from the Raster Functions pane or on disk. Once you choose the processing template, ArcGIS AllSource verifies whether the template is valid.
You can use processing templates that require parameter values to be set for the processing to work. Once you choose a template, you may get a notification indicating that the template is invalid and the OK button is disabled until you validate the template or choose a valid template. Use the Inputs section to set the values of the required parameters. You can use the Add inputs to the current template button 
to change multiple parameters in the raster function template processing chain. Click the Validate button 
to revalidate the template. Once the template is successfully validated, the OK button is enabled. Click OK to apply the template to the mosaic dataset or image service layer.
Export and download data
When working with mosaic datasets, you may eventually need to create a permanent output of an analysis. With the exception of geoprocessing tools, all of the analysis you perform on a mosaic dataset or an image service occurs in memory. The Export Data button 
opens the Export Raster pane, where you can write a raster to disk. If you do not want to download the source files from a mosaic dataset or image service, you can save the image that is displayed, or a part of it, using the Export Data option.
In addition to saving data to disk, you can use the Download Rasters button 
to open a pane where you can download the source data from an image service and mosaic layers. This accesses the service and moves the actual files onto your machine, rather than reading the compressed version.
The downloading capability is controlled by the administrator of the image service, and not all image services allow you to download the source files. You can choose the desired source files using the attribute table or one of the selection tools. The source files may include raster datasets, LAS files, or LAS datasets. Once selected, any additional associated files will be included, such as metadata files or projection files. From the Select drop-down menu on the Data tab, you can draw an area of interest, make a selection based on an attribute, or select only the visible rasters. If you only want to download the source rasters, choose Deselect Overviews using the Operations button 
. These options are also available from the layer's context menu in the Contents pane. When the selection is ready, use the Download Rasters button to proceed, and complete the following steps:
- Type a path or click Browse
to specify a
location on disk where the data will be downloaded, and check the items in File List that you want to download. If the row check box is not checked, the item will not be downloaded. - Optionally, clip the data based on the extent of the data frame or any of the layers in the Contents pane.
- Check the Convert Raster check box to set the parameters for how the output is saved, including the file format and compression method.
- Optionally, check the Add output to map check box to display the downloaded data.