Re: Compression and PSNR

> @But what you *can* do is to measure the rate/distortion curve of
> both codecs, even though the sample points are not identical. You
> should be able to interpolate between the two, at least graphically:
> Plot PSRN against BPP and compare the plots.

> But if you think about it, you are looking at the ratio I mentioned in
> my first post even in the curves. Slope, which would define which
> method is better is infact this ratio. Nonetheless I think that this
> method is better.

This actually is what I mean. (-; By looking at the plots, you can
also derive a region where the codec at hand performs better (low/high
bitrate) as well as in absolute numbers (rate larger/smaller) as well
as differentially (looking at the slope). The slope will likely not be
constant along the curve either.

> @If you really got 1000 people, and you setup the test correctly
> (i.e. you haven't tried to bribe them ;-) then this sounds pretty
> much convincing to me.

> But how would I prove this to the editor?

Well, I would describe the experiment and deliver some of the outcomes
in numbers. As in, "we asked 1000 xxx-students to participate in our
test and rate the streams according to a scale from ... to ....
points. We instructed them to check for .... According to figure xxx,
about xxx% of the students rated... We found artifacts of the kind xxx
in streams xxx" and so on. Describe the methods accurately, give the
statistics. Do not mix your conclusions with the description of the
experiment, state facts first, then derive results, make clear what's
the experiment and what conclusions you draw from it.

So long,
Thomas
.

Relevant Pages

  • Re: points on a trendline
    ... If it's a straight line, you can use INTERCEPT and SLOPE ... you can use LINEST according to Bernard ... >> Some of the plots are not very smooth so I added a trend line. ...
    (microsoft.public.excel.charting)
  • Re: Buying land: What should you look for?
    ... > For the plots I'm considering, the local power lines either border the ... > slope would hide the house from the road. ... > my previous gardens have been on southern slopes. ... not me - still very slowly terracing the parts that aren't ...
    (misc.rural)
  • Re: Buying land: What should you look for?
    ... For the plots I'm considering, the local power lines either border the ... The property tends to slope north (with a few gullies ... my previous gardens have been on southern slopes. ... anyone had any experience in terracing wooded country? ...
    (misc.rural)
  • Re : cubic equation - is it really solvable?
    ... >>> my excel program solves all cubics and plots the curves without any ...
    (sci.math)
  • Re: "Innaccurate" multiplot?
    ... several curves into one "graph". ... Multiplots can usefully overlap, but that needs special care or fulfills special needs. ... Or, with a good deal more trickery, you can actually overlay multiple independent plots in the same region. ...
    (comp.graphics.apps.gnuplot)