Raster images and surfaces

I'm sure you are all familiar with the seminal post by Justin regarding Raster Troubleshooting.  I thought I might add in a few more thoughts on factors that can impact raster usability, especially now that 64-bit operating systems are becoming prevalent. 

• FDO vs. Insertion: An inserted image (via ImageAttach, MAPIINSERT, or IINSERT) is similar to an XREF in the way it behaves – on opening a drawing that contains an inserted image, the entire thing will be scanned in, which can really slow opening time.  An FDO connected image (DATACONNECT or MAPCONNECT, then the FDO Raster Provider) is just linked by a pointer to the file, and only the necessary area for display will be polled/queried and displayed.
• Number of pixels: there’s no hard and fast limit, but any time that you can keep your image below a billion (1,000,000,000) pixels you’ll have better luck, especially for inserted images. FDO can handle large images more efficiently.  You’ll hit a billion pixels faster than you expect – one square mile at 6 inch resolution (6 inches per pixel) is about 111 million pixels.  Bump that resolution up to 3 inch, and you are at 446 million pixels.  That might sound like really high resolution, but remember that LiDAR posting can be about that resolution, so creating a surface at that scale could easily be attempted.
• Pyramid layers: these are images that are multiresolution, such as MrSID.  That means that if you imagine the different resolution layers as forming a pyramid, the original, full scale imagery is the bottom layer.  As you move up, each layer contains fewer and fewer pixels, but still gives enough information at the right scale to determine what you are looking at, until you get to the top layer, which is basically a thumbnail of the base layer.  It would look terribly pixelated zoomed in close by itself, but for the view from space, it’s just fine.  Another fairly common multiresolution format is JPEG2000, and some software packages support external pyramid layers, which are generally saved as .aux files with the same file name.  Generally, when printing there’s no benefit to pyramid layers, as the bottom layer will be referenced, regardless of the layer displayed on screen.
• Compression: this can be a double edged sword, as good compression can drastically decrease the size of the image while maintaining visual fidelity, but in some instances, like printing, the computational overhead of decompressing can be significant. If you are having trouble, it’s always worth saving an image out as an uncompressed tif to see if the behavior changes.
  • As for the type of compression, you have lots of options: both the MrSID and JPEG2000 options above offer compression in addition to pyramid layers. 
  • ECW has a great reputation for their compression algorithm and its performance.
  • JPEG compresses very nicely, and even TIF images can be created with compression.
  • Important note: even with Raster Design, you can’t save to MrSID or ECW formats due to their proprietary nature (you can insert or connect to all of them).  Most free image software, such as Irfanview or XNView will allow you create uncompressed tifs, JPGs, and many other file formats.  If you look, there are even free tools that will do size limited exports of the proprietary formats and open source projects for other formats.  Those, and the free tools referenced above, are, of course, use at your own risk.
  • FDO imagery has an option to refresh that is independent of REGEN – right click the layer in the Display Manager to get the most appropriate display for your view scale.

Have fun!

 

(thanks to MassGIS for the image above)