home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.compression:3122 sci.image.processing:574
- Newsgroups: comp.compression,sci.image.processing
- Path: sparky!uunet!snorkelwacker.mit.edu!news.media.mit.edu!news.media.mit.edu!popat
- From: popat@image.mit.edu (Kris Popat)
- Subject: Re: Best compression for 24 bit images
- In-Reply-To: davidsen@ariel.crd.GE.COM's message of 28 Aug 92 13:35:52 GMT
- Message-ID: <POPAT.92Aug28140827@image.mit.edu>
- Sender: news@news.media.mit.edu (USENET News System)
- Organization: MIT Advanced Television Research Program
- References: <1992Aug28.133552.16085@crd.ge.com>
- Date: Fri, 28 Aug 1992 19:08:27 GMT
- Lines: 32
-
-
- > What is the best way to do lossless compression of 24 bit images? I am
- > currently using 24 bit tiff with lzw and the pnmtotiff program
- > (-predictor 2). JPEG is not lossless and can't be considered, the
- > restored image must be pixel by pixel identical.
-
- JPEG does have a lossless coding mode -- you might want to see how it
- compares in performance with what you're doing now.
-
- Not a whole lot of research has been done in lossless compression of
- images, and of the work that has been done, the compression ratios are
- not all that dramatic (e.g., 2.5:1 on average). The most popular
- techniques among researchers seem to be variations on bit-planing in
- conjunction with adaptive binary entropy coding.
-
- One of the limiting factors is camera noise -- which can require as
- much as one bit per pixel (yes, really), and is inherently
- uncompressible. Thus the largest compression factors we can ever
- expect to see are in the range of 8:1 for typical 8-bit monochrome
- natural-scene images. Any technique that consistently provides more
- than 3.5:1 should be considered a real breakthrough. Such a
- breakthrough is likely to happen over the next few years, as there is
- no fundamental reason why 4:1 or 5:1 can't be done with a
- sophisticated enough algorithm.
-
- Note that these numbers are for typical CCD cameras and typical
- resolutions, e.g., 512x512, using natural scenes. We could of course
- get much higher compression ratios by starting with oversampled,
- low-noise original images.
-
- Kris Popat
- MIT Rm. E15-355 Cambridge, MA 02139
-