home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!warwick!uknet!strath-cs!ksaetre
- From: ksaetre@cs.strath.ac.uk (Kjetil Saetre IE90)
- Newsgroups: comp.graphics
- Subject: Contours
- Keywords: Reconstruction, Coding
- Message-ID: <11300@baird.cs.strath.ac.uk>
- Date: 14 Dec 92 13:30:22 GMT
- Sender: news@cs.strath.ac.uk
- Organization: Univ. of Strathclyde, Glasgow, UK
- Lines: 26
- Nntp-Posting-Host: fleming-10
-
- As part of my 4th year university project, I have some tasks
- where I need some ideas on hove how to improve things.
-
- (A)
- I must be able to reconstruct an grey scale image from some
- of its contours (e.g. 10). Until now I have used a ramp function
- sliding the colour from one contour to the next horisontally. This
- often gives errors in images where the contrast is high. In such images
- many contour levels will be laid on top of each others and information is thus
- lost. Other methods ? Please write.
-
-
- (B)
- Some input on coding schemes of contours will be highly appreciated. By now
- I have used Huffman variable length coding with some success. I have also
- used Bezier approximation (with bernstein polynomials) and I have tried
- a Fourier approximation. I still can not manage to get the representation
- size of the contours below 1.5 bits/pixel. I have heard about others doing
- much better.
-
- Note: the 1.5bits/pixel is obtained in an 256x256 image using 256 greylevels and is
- a picture taken in an underground station. Typical contour length 3000 pixels.
-
- If you know of anything that may help me please reply.
-
- Kjetil Setre
-