L1 - Median (a multivariate median)

There are several papers that give algorithms that are described as
"simple" and "rapidly converging", however, I find that naive
implementations are not robust to finite precision arithmetic. For
example, a test such as "if(x==0)" must be replaced with "if(x<epsilon),
where epsilon is a small number.

Making the modifications to deal with finite precision arithmetic has been
a real headache. Does anyone have working source code that I can examine?

Incidentally, in case you are wondering, I am using this to remove passing
objects from stacked images.

.

Relevant Pages

  • Re: Help please..
    ... likely you will find faster general-purpose implementations ... than LAPACK on general-purpose platforms. ... compile the source code for your target. ... ESPRIT algorithms some ten years ago. ...
    (comp.dsp)
  • Re: Timings issue
    ... >> You mean to say that timings of implementations of cryptographic ... >>algorithms on different platforms are not relevant to cryptography? ... who build high speed code are also experts in how to design code that ...
    (sci.crypt)
  • Re: Please help me find a tagline that doesnt suck
    ... provide clients worldwide with robust, scalable, modern turnkey ... implementations of flexible, personalized, cutting-edge Internet-enabled ... e-business application product suite e-solution architectures that ...
    (comp.lang.lisp)
  • [OT] Re: C Books and algorithm books
    ... reading several pages, feeling I was so intelligently inferior. ... quantitative comparisons of implementations. ... You may have heard that Quicksort and Heapsort are both O ... algorithms; this much needs only a rather loose ...
    (comp.lang.c)
  • Re: Irrational numbers
    ... > You've already been given the names of the algorithms and the ... I want the implementations so that I could try out my ... >> one could determine the unknowns or not. ... the 'integer relation' to be furnished by Percival ...
    (sci.crypt)
Quantcast