2001.04.26 18:34 "8K chunks for RowsPerStrip?", by Chris Barker
I suggest tiling is much more appropriate for large images.
Tiling can (especially for 2-D compression schemes have better compression than one piece images, and is never *much* worse. It tends to isolate areas that negatively compress into one (or a small number) of tiles.
Unfortunately, the current TIFF spec does not provide for tile-types (compressed, uncompressed, foreground color, background color), but still tiling works very well, and is much more effective in limiting memory usage. Of course, memory is cheap now, but small devices still have limited memory. Also, some versions of windows have trouble with large bitmaps, whether whole or striped, especially when trying to enlarge them onto a high resolution print page.
Striping makes rotating and scaling images (without combining them into one large one) difficult, but suitable square tiles have no such problem. The only difficulty is that few implementations support it.
Suitable tile sizes are 32K for bitonal and 64K for color.
Does this standard still apply? I end up seeing _lots_ of TIFF files that write out one strip. Seems like this standard/suggestion (per Section 7 of the spec) was probably meant more for days when machines didn't have much ram and swap space.
I've also noticed that most apps I can find have no trouble reading one strip images and one strip images, of course, have better compression. Sometimes _significantly_ better compression.
Barker's law: For every inaction, there is an equal and opposite excuse.
Check out Interlinear's web page at http://www.ilt.com