Re: Question: Research without measurement



Interesting set of replies. It's true that it can be very expensive to
test the usability of a computer system, or the productivity of
programmers using certain tools. But, there are other areas where I
feel that there is less justification. E.g. someone might implement an
image processing algorithm where the expected result is that the
processed image will look better to humans. How hard can it be to apply
it to a large number of images, then set up some sort of web page that
serves randomly ordered processed and unprocessed images and asks which
is better. It may be impossible not to know which is processed and
which is unprocessed leading to the viewer knowing which answer is
expected. Even then the design of the image processing algorithm can
still be validated by blind comparison to alternative algorithms, or
simplified versions of the proposed algorithm so that the full
complexity is shown to be necessary. There have been some very good
examples of cases where experimentation can be difficult and/or
expensive mentioned above. But IMHO there is a lot of "computer
science" where robust experimentation is far, far easier and cheaper
than other sciences.

Note: I'm not claiming that people working on Image Processing
typically research in this way. But, it was one Image Processing paper
that someone showed me that had been published in a good journal that
got me thinking about this.

Cheers,

Ross-c

.



Relevant Pages

  • Re: Cpython optimization
    ... This would be way to speed up things in an image processing algorithm: ... I'm sure that performance-conscious image processing software already has ... Give Ulrich Drepper's "What Every Programmer Should Know about Memory" ... to speed up array ops. ...
    (comp.lang.python)
  • Re: image processing algorithm
    ... what is actually image processing algorithm and how to describe it? ... where N and M are the dimensions of the image starting at zero. ... Reduce the focus on me- by being really clever in front of myself - And not ...
    (comp.dsp)
  • Re: image processing algorithm
    ... what is actually image processing algorithm and how to describe it? ... made up of discrete pixels - those pixels usually made up of binary words or ... where N and M are the dimensions of the image starting at zero. ...
    (comp.dsp)
  • Re: zernike moments question
    ... On Mar 8, 8:26 am, Adam Chapman ... I have to calculate the zernike moments (http://homepages.inf.ed.ac.uk/ ... an image processing algorithm. ... What "image processing algorithm?" ...
    (sci.image.processing)