2007.01.17 02:25 "[Tiff] Elevation Data", by Craig Bruce

2007.01.17 21:18 "Re: [Tiff] Elevation Data", by Frank Warmerdam

In my (prejudice, on-track and very limited) mind, large image handling is pipeline-based. If I need to render a large TIFF (or any TIFF for that matter), I set up a pipeline that decodes, resamples, and converts color and all. I next start pulling the pipeline, and as results come in, I progressively update display.

That's just one very typical example, of course, of what I regard 'progressive streaming' of imaging processes. But it's sufficient to explain the dynamic sampling is rather awkward in this situation. I'd have to set up a first pipeline that decodes and gathers range, and run the complete pipeline, before I'd be able to set up the second pipeline with the detected range in the color conversion steps. For huge images, that's many seconds before the second pipeline is even build, which conflicts with my need to start progressive build-up of display asap.


OK, I see your point. when I saw progressive streaming I wondered if we had morphed into a talk about reading stream data from the net or something.

For the record, I only *sample* the data to establish a range for scaling. I have an algoirthm that takes roughly the square root of the number of tiles/strips, reads them and uses that as a presumed reasonable range to scale. This takes relatively little extra time even for very large images.

Best regards,

I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam@pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam

and watch the world go round - Rush    | President OSGeo, http://osgeo.org