LibTiff Mailing List
TIFF and LibTiff Mailing List Archive
January 2009
Previous Thread
Next Thread
Previous by Thread
Next by Thread
Previous by Date
Next by Date
Contact
The TIFF Mailing List Homepage
This list is run by Frank Warmerdam
Archive maintained by AWare Systems
Thread
2009.01.13 21:01 "Re: Libtiff compression ratios", by <myopicwatchmaker@gmail.com>
Setting TIFFTAG_PREDICTOR to PREDICTOR_HORIZONTAL worked beautifully. With the predictor running, libtiff's compression ratios reach parity with Photoshop. Increasing the deflate compression ratio to its maximum via TIFFTAG_ZIPQUALITY had only marginal effects on compression (a 2% improvement) while tripling the computation time required. I appreciate the advice! Perhaps PREDICTOR_HORIZONTAL might make a good default value for the libtiff predictor tag? Nick Chris Cox wrote: > Photoshop may work on larger strips than LibTIFF. > And Photoshop does use the horizontal predictor by default with LZW and > Flate compression. > > Chris > > > > > On 1/11/09 5:19 PM, "myopicwatchmaker" <myopicwatchmaker@gmail.com> wrote: > > Hi All, > As part of my doctoral thesis work, I produce two and a half terapixels of > grayscale images a month. As you might imagine, I am eager to minimize > the storage requirements for this data. Libtiff's support for deflate > compression has served me quite well up until this point. > > Recently I noticed that with identical compression settings, Adobe > Photoshop CS2 achieves significantly higher compression ratios than my > libtiff code--producing files that are almost 40% smaller in many cases. > > A typical example: I have a 30,000x6500pixel 8-bit grayscale image which, > uncompressed occupies 174Mb on disk. Libtiff's deflate algorithm > compresses this down to a 66Mb file, where Photoshop produces a 48Mb file. > I save my images in ~8Mb strips, which appears to be the optimum for zip > compression. > > Does anyone know why libtiff might fail to match Photoshop's compression > rates? Are there any paramters in libtiff or libzip that might allow > optimization for humungous images? > > Thanks for any input. > Nick > >