LibTiff Mailing List
TIFF and LibTiff Mailing List Archive
November 2006
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
2006.11.25 18:41 "Re: Windows HD Photo - any interest?", by Sachin Garg
On 11/25/06, Andrey Kiselev wrote: > On 11/24/06, Brad Hards wrote: > > Not sure if this has been discussed before, but Windows Vista comes with a new > > image format (previously known as Windows Media Photo, now called HD Photo). > > BTW, I built the tools on my Linux box (surprizingly painless, just > single fix in sources!). I played with lossless mode of the encoder > and found that on my images (RGB photos) it is either comparable or > better than Deflate with horizontal differencing predictor. In all > cases HDPhoto files were smaller up to 10%. It wouldn't be fair to compare something state-of-the-art (like HDPhoto) to decades old Deflate :-) I haven't taken HDPhoto for a test ride myself, but I once compared PNG (which is almost same as your deflate+predictor), Jpeg-ls and Jpeg-2000 for lossless image compression of RGB images. I can share those results, you might find them interesting. Jpeg-ls is consistently 2-3% better than jpeg-2000 lossless. And Jpeg2000 lossless is around 7-8% better than PNG. Combining this with your results, it means HDPhoto will be somewhere close to jpeg-ls. Also, Jpeg2000 and PNG take almost same time and jpeg-ls is almost 3 times faster than both. Do you have results on HDPhoto's speed compared to deflate? That will help complete the picture. This is not a very reliable way to compare formats, but still might give an insight. If someone else has performed similar tests, I would love to hear what they got. Sachin Garg [India] www.sachingarg.com | www.c10n.info (For jpeg-ls, I had applied lossless RGB-to-YCbCr conversion first, this is not a part of standard but it makes sense to do this as standard recommends doing such channel de-correlating transforms beforehand. Without this, jpeg-ls gives results almost same as PNG. Jpeg2000 inherently does this same conversion. I had used kakadu for jpeg2000, University of British Columbia code for jpeg-ls and libpng for png.)