Available with Standard or Advanced license.
When you need to recalculate the footprint to remove those pixel values you don't want in your mosaic dataset, you can use the Build Footprints tool. For example, your footprint is originally the entire dimension of the image, but because it was rotated, you want it to reflect the actual valid pixel values of your images. 
There are a number of options on the Build Footprints tool that can be modified that will affect the footprint output. For example, there may be a range of pixel values on the high and low ends of your data values that you don't want included. Normally, these pixel values can easily be defined, but if the data is stored with lossy compressions, the values will be a bit fuzzy. For example, instead of all values of 0, your values may be 0–3.
Recommended parameter settings
The table below defines the various parameters that can be altered to radiometrically adjust the footprints.
| Parameters | Description |
|---|---|
Minimum data value | Exclude pixels with a value less than this number. For example, with 8-bit data, the values can range from 0 to 255. A value around 0 represents very dark colors, like black border pixels. When you specify 1, then the only value less than 1 is 0, so all 0 values will be considered invalid data and will be removed from the perimeter of the footprint. If the imagery is compressed using a lossy compression method, then you should define a value slightly greater than 1 to remove all the black pixels. When dark areas, such as shadows, have been incorrectly removed from the footprint, this value should be decreased. |
Maximum data value | Exclude pixels with a value greater than this number. For example, with 8-bit data, the values can range from 0 to 255. A value around 255 represents very bright colors, such as white clouds and snow. If you specify 245, then all values between 246 and 255 will be removed from the perimeter of the footprint. |
Approximate number of vertices | Choose between 4 and 10,000. More vertices will improve accuracy but can extend processing time. A value of -1 will calculate all vertices. More vertices will increase accuracy but also the processing time. The minimum value is 4 and the maximum value is 10,000. The greater this value, the more accurate and irregular the polygon and the longer the processing time. A value of -1 will show all the vertices in the footprint, therefore your polygon footprint will not be generalized. |
Shrink distance | Clip the footprint by this distance. This can eliminate artifacts from using lossy compression, which causes the edges of the image to overlap into NoData areas. Shrinking of the polygon is used to counteract effects of lossy compression, which causes edges of the image to overlap into NoData areas. |
Request size | Set the resampled extent (in columns and rows) for the raster when building footprints. Greater image resolution provides more detail in the raster dataset but increases the processing time. A value of -1 will compute the footprint at the original resolution. You can increase or decrease this value based on the complexity of your raster data. Greater image resolution provides more detail in the raster dataset and thereby increases the processing time. A value of -1 will not resample the footprint, therefore it will compute the footprint at the native pixel size. |
Minimum region size | Avoid small holes in your imagery when using pixel values to create a mask. For example, your imagery may have a range of values from 0–255, and to mask clouds, you've excluded values from 245–255, which may cause other, non-cloud pixels to be masked as well. If those areas are smaller than the number of pixels specified here, they will not be masked out. This value is specified in pixels, and it is directly related to the Request Size, not to the pixel resolution of the source raster. |
Maintain sheet edges | Check when using raster datasets that have been tiled and are butt joined (or line up along the seams with little to no overlap). An analysis of the image edges is performed so that the sheet edges are not removed. |
The tables below define various types of datasets that sometimes have NoData borders and the recommended parameter settings.
Regular and clipped to sheets
Used for datasets where the data pixels in each image form a rotated rectangular area. The resulting imagery will be clipped to a new sheet or tile. Such datasets are typically created by reprojection of images or scenes, then cut to map sheets or tiles with little or no overlap. The parameters are set such that the footprint will contain only a few vertices. An analysis of the edges is performed to maintain the sheet or tile boundaries.
| Parameter | No or lossless compression | Lossy compressed |
|---|---|---|
Minimum data value | 1 | 10 |
Maximum data value | 254 | 245 |
Approximate number of vertices | 4 | 4 |
Shrink distance | 2 pixels | 6 pixels |
Request size | 1000 | 1000 |
Maintain sheet edges | Yes | Yes |
Irregular and clipped to sheets
Used for datasets where the data pixels do not form rectangular areas. The resulting imagery can be cut into map sheets or tiles. Such datasets are typical for imagery along pipelines or other linear features. Here, the pixel areas cover the linear feature, and the images are then mosaicked and cut into tiles. The parameters are set to allow a larger number of vertices to define the border. An analysis of edges of the resulting footprint is performed to maintain the sheet or tile boundaries.
| Parameter | No or lossless compression | Lossy compressed |
|---|---|---|
Minimum data value | 1 | 10 |
Maximum data value | 254 | 245 |
Approximate number of vertices | 30 | 30 |
Shrink distance | 6 pixels | 6 pixels |
Request size | 1500 | 1500 |
Maintain sheet edges | Yes | Yes |
Rotated rectangle
Used for images that form a rotated rectangle. Such datasets are typically created when individual scenes or map sheets have been rotated and the sides of the footprint remain straight. The parameters are set to define the footprint only by four vertices.
| Parameter | No or lossless compression | Lossy compressed |
|---|---|---|
Minimum data value | 1 | 1 |
Maximum data value | 254 | 245 |
Approximate number of vertices | 4 | 4 |
Shrink distance | 2 pixels | 6 pixels |
Request size | 1000 | 1000 |
Maintain sheet edges | No | No |
Reprojected rectangle
Used for rotated images that have been reprojected, which form rectangular footprints with curved edges. Such datasets are typically created when individual scenes or map sheets have been reprojected. The parameters are set to define the footprint with a sufficient number of vertices to represent the curves.
| Parameter | No or lossless compression | Lossy compressed |
|---|---|---|
Minimum data value | 1 | 10 |
Maximum data value | 254 | 245 |
Approximate number of vertices | 20 | 20 |
Shrink distance | 2 pixels | 6 pixels |
Request size | 2000 | 2000 |
Maintain sheet edges | No | No |
Orthorectified image in flat terrain
Used when the origin of the raster dataset is a scene or image that has been orthorectified to a flat terrain. The edges of such images form simple curves caused by smooth changes in the elevation.
| Parameter | No or lossless compression | Lossy compressed |
|---|---|---|
Minimum data value | 1 | 10 |
Maximum data value | 254 | 245 |
Approximate number of vertices | 30 | 30 |
Shrink distance | 9 pixels | 9 pixels |
Request size | 2000 | 2000 |
Maintain sheet edges | No | No |
Orthorectified image in hilly terrain
Used when the raster dataset is being orthorectified in an area that encounters large changes in elevation. The edges of such images are irregular, caused by sudden changes in the elevation. More vertices are required to define such footprints.
| Parameter | No or lossless compression | Lossy compressed |
|---|---|---|
Minimum data value | 1 | 10 |
Maximum data value | 254 | 245 |
Approximate number of vertices | 200 | 200 |
Shrink distance | 9 pixels | 50 pixels |
Request size | 2000 | 2000 |
Maintain sheet edges | No | No |
Minimum region size
A suggested default for this parameter was not specified in the tables above because of the potential complexity of this value. You need to consider what the raster data is and how detailed you require the footprint to be.
Your footprint is designed to define the area of the raster dataset you want to view. The footprint is typically the extent of the raster dataset; however, it can be modified so a user cannot view part of the raster dataset.
The minimum and maximum data values are used to specify the valid data. Outside those values is a range of pixel values that will be used to create a region, or continuous feature, in the raster dataset that will create a hole in the footprint polygon. For example, if you have a raster dataset with pixel values from 0 to 255, you can define the valid range as 10–255. Therefore, the pixels from 0 to 9 will result in holes in the footprint. However, your image may have pixels from 0 to 9 that you want to retain because they represent valid features. You need to consider what these may be. For example, if the rooftop of a home is valid but is likely to have pixel values from 0 to 9, you need to ensure these do not result in holes in the footprint; whereas larger features, such as large clouds, do.
The area of the hole is compared to the area computed using the minimum region size. If the area of the hole is smaller than the area computed by minimum region size, the hole is removed. Minimum region size ensures that only reasonably large features are eliminated from the footprint by having them remain as holes in the computed geometry. All candidate holes that are smaller in area than that area denoted by this parameter are removed; that is, they no longer appear as holes in the computed footprint. thus ensuring such tiny features are not clipped from the dataset.
In the diagram below, the gray pixel values represent the valid data values. The orange pixels represent three regions in this raster containing values you potentially want excluded. The yellow boxes represent the area defined by the minimum region size. Since the two small orange regions on the left are smaller than the minimum region size (yelllow), the pixels are not excluded. However, the large orange region on the right side is larger than the minimum region size, therefore those pixels are excluded from the mosaicked image in the mosaic dataset.
The request size also needs to be considered when specifying the minimum region size, because the request size determines the resolution or detail in the raster used to recompute the footprint. Normally, you will use a request size that is smaller than the dimensions of your original source raster. This will impact your region size. For example, if you want to preserve the rooftop area, you need to consider the roof's pixel dimensions in the source raster and its dimensions in the request size. Therefore, if the roofs are 50 x 50 pixels in the source raster and the request size represents a raster that has 50 percent fewer pixels in its x and y dimensions, the rooftops may be represented by a feature that is only 25 x 25 pixels in dimension. You need to be cognizant of the request size versus the minimum region size so you don't exclude features you want to retain but still exclude any desired features by creating holes in the footprints. In other words, you need to define the correct size to remove the small holes but retain the larger holes.