2007.07.03 18:37 "[Tiff] BigTIFF extension?", by Phil Harvey

2007.07.04 17:49 "Re: [Tiff] 16-Bit-Per-Channel Lossless Compression", by Stephen Carlsen

Hi Andy,

Yup, I've actually done quite a bit of testing, and running LZW with predictor tends to compress 16-bit digital camera photographs very little, or even not at all. The problem, of course, is that the low order bits tend to look a lot like noise.

I can post some sample files if you like. On one of my images from my Canon XTi, exported from DNG to uncompressed TIFF (using Lightroom), the uncompressed size is 57.7 MB, but is 68.8 MB with LZW (plus predictor) compression.

Flate tended to generally do worse than LZW, and be a lot slower, in my testing.


PS With LZW, have you tried with prediction enabled? That should improve the compression ratio on 16 bit ints.

From: "Andy Cave" <andy.cave@hamillroad.com> To: <tiff@lists.maptools.org>; "Stephen Carlsen" <sc42business@mac.com> Sent: Wednesday, July 04, 2007 6:24 PM Subject: Re: [Tiff] 16-Bit-Per-Channel Lossless Compression

Are you saying that LZW or Flate do not get 2:1? Have you tried either of these? What ratios do you get?

Do you have some example files for testing with?

I think if you can compress and write to disk faster than you can write, compression is always worth it. Sure storage space is getting larger and faster, but then CPUs are getting more and more cores, which can nicely do compression in parallel.

It seems to me that none of the lossless compression schemes commonly used in TIFF are very effective at compressing 16 bit or 32 bit data. And, with technologies such as Camera Raw gaining in popularity rapidly, there's a lot more 16 bit data out there.

If we can come up with a simple approach that gives us on the order of 2:1 compression of 16 bit photographs, is this useful, or are storage space and bandwidth now so inexpensive that we don't care any more?