home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.compression.research
- Path: sparky!uunet!mcsun!sunic!aun.uninett.no!nuug!nntp.nta.no!hal.nta.no!kol
- From: kol@hal.nta.no (Karl Olav Lillevold FST Student)
- Subject: Bits needed to code image block?
- Message-ID: <1992Sep11.062036.25959@nntp.nta.no>
- Sender: news@nntp.nta.no
- Nntp-Posting-Host: logos.nta.no
- Organization: Norwegian Telecom Research
- Date: Fri, 11 Sep 92 06:20:36 GMT
- Lines: 23
-
- I am not very experienced in validating compression algorithms, and
- I now need to estimate the number of bits needed to code an image block
- of an error image. The error image is the result of a motion
- compensation, and I assume the coding procedure will be something
- DCT-like, i.e. blockwise coding.
-
- I have one advice: Find the squared error of each block and subtract
- the gain obtainable by removing correlation. This should (somehow)
- give me an estimate of the needed number of bits.
-
- My questions are:
- 1) Does anyone have a practical solution to this problem?
- 2) Is there a common procedure that is normally used so that
- different algorithms can be compared without needing to
- do the actual coding? (I have seen results based on
- both SNR and entropy)
-
-
- Karl Olav Lillevold email: kol@hal.nta.no
- Student Norwegian Institute of Technology (electronics)
- Diploma at Norwegian Telecom Research (image processing)
-
-
-