1998.09.22 14:51 "Photometric interpretation of multi-band, NON-alpha images ?", by Ed Grissom
This message is one of a series of messages with questions about advanced raster topics and TIFF. See the message entitled "TIFF Questions - TIFF 7.0 ?" for more info.
- Photometric interpretation of multi-band, NON-alpha images ?
Typically, satellite data vendors get their data from the downlink and convert it to a format that is useable by their customers. They may offer a variety of formats for the data so that the customer can choose whatever format suits their needs. At least one satellite data vendor uses our software to create some raster data output formats for sale to their customers. It is possible that these data customers also use our software to exploit the imagery for their own purposes.
The data vendor would like to have a single output format that most or all of their customers could use. TIFF seems to be the best choice for this, given its flexibility, widespread use, and the capability of GeoTIFF to internally specify what portion of the earth is covered by the image.
Current Landsat imagery has 7 bands; new satellites will routinely have at least 4 bands of valid image data and some may have 20 or more. We would like to be able to convert any type of satellite data to TIFF to take advantage of tiling and overviews which help with performance during any processing activities. My concern is not with the reading of the raw satellite data, it is with getting it into a reasonable TIFF format.
The current definitions for the Photometric Interpretation tag seem to be lacking an entry that applies to multi-spectral imagery. Imagery gathered from satellites typically have sensors that respond to several frequencies of visible or microwave light. These frequencies do not necessarily correspond to "RGB", "CMYK", "YCbCr", or "CIELab".
However, when viewing these images, it is common to generate a "false-color" image by using some combination of the bands as RG&B. For instance, viewing vegetation might be done by combining bands 4,5, and 1 of Landsat data as R,G, & B respectively. For viewing soil moisture, a different arrangement of three of the seven bands might be used. It is also common to just look at one band at a time as a greyscale image.
In looking at the TIFF spec for the correct way to store these separate images, I came up with several possibilities: ("+"=Pros, "-"= Cons)
1) Store as separated (CMYK) images.
+ Allows multiple bands, not limited to 4.
+ Allows "InkNames" tag which could be used to document the sensor frequency band.
- Requires data to be inverted from intensity to "%ink".
- Not recommended for general image interchange.
+ TIFF/IT defines additional TAGs that might be useful here, although that would be using the TIFF/IT spec for something that it was not intended to be used for.
- TIFF/IT support requires quite a few other capabilities that are not necessary or are poorly defined for these images.
2) Store as "MinIsBlack" image with ExtraSamples Tag set to have numband-1 entries, each set to "Undefined".
+ Does not seem to be illegal.
- May choke 3rd-party apps that do not support ExtraSamples.
- May not be widely adopted even in our industry without "official" support.
3) Store as separate GreyScale images in a multi-image file (linked list of IFD's)
+ Least impact to current readers, most either handle or ignore multiple IFD's.
- Greatest impact to applications that want to use this imagery as RGB bands, since separate IFD's are treated as unrelated images.
- GeoTIFF info must be repeated for each IFD.
4) Lobby for specific "multi-spectral" related enhancements in TIFF 7.0
+ Will be a standard that can be used for general interchange
- Does not appear to be a 7.0 standard forthcoming.
- Adobe does not appear to be interested in further developments.
Any suggestions on the best route to go here???