PromptServer is a geospatial image server that I found out about when reading the GIS Monitor newsletter. The approach to managing imagery that was described in the article was quite interesting so I did some quick web research.
Disclaimer: What follows is based solely on the information available from the www.promptgeo.com website. Your reality may vary. (No, this post is not an April Fools joke.)
Map servers can be divided into two groups:
PromptServer is a server-side renderer, but it differs significantly from other products in that group.
Due to the size of geospatial image files, the typical approach to preparing imagery for serving by a map server is:
- Obtain all the required image data.
- Use image processing software to create a physical image product. e.g. A seamless mosaic or tiles that form a mosaic.
- Archive the original data, and any useful intermediate products.
- Store the image product in a file system or database, ready for use by the map server.
- Create the metadata that describes image product. (This should be step no.3, but creating metadata always seems to be done last, if at all.)
Some map servers can do some basic value added processing to that image product before they serve it, like reprojection on-the-fly, but that’s about it. PromptServer is quite different in that you don’t need to create physical image products, and then store and manage those products. It allows you to build up catalogues of georeferenced imagery and create “virtual image products” on-the-fly from those catalogues.
In essence, PromptServer is an image processing engine and map server in one package (that’s my interpretation). Virtual image products are created by defining “processing chains”. Each request to the server invokes a processing chain that creates a map image. (It reminds me of the algorithm approach used by ER Mapper.) PromptServer plug-ins are available for:
- Band processing for multispectral imagery
- Clipping
- Histogram statistics extraction
- Enhancement using look up tables and filters
- Pan sharpening
- Warping
- Orthorectification
- and more…
There are several data management benefits to using the virtual image process:
- You only need to store the original data. No intermediate or final data products physically exist. Less data means easier data management.
- As more data becomes available should be easy to add it to the appropriate catalogue and make it available to the applications that use it.
Image processing is a CPU, memory and disk intensive activity. PromptServer’s response times are going to depend on the type of image data, the processing that needs to be applied to it and the server hardware is running on. I’d be interested to know what sort of hardware would be required to generate typical image within a reasonable period (less than 5 seconds). PromptServer is certainly positioned well to take advantage of the ever reducing cost of CPU cycles, RAM and disk space.
Footnote: I have sent an email to the good folks at PromptGeo to let them know about this post, and offering them the opportunity to provide feedback or leave a comment. If I’ve got anything wrong I’ll change the post.