Re: Calculation of snr

On May 23, 9:41 am, "David J Taylor" <david-tay...@xxxxxxxxxxxxxxxxxxx
this-bit.nor-this-bit.co.uk> wrote:

Don,

It sounds as if we have dealt with quite similar things in the past! We
used a technique of taking a four-bar target at a particular contrast
(delta-T) and spatial frequency, and applying /all/ the MTF and noise
factors to that (atmosphere, lens, sensor, processing, display, eye/brain)
and working out the SNR on the retina. Subjective tests showed what
actual observers could achieve in terms of subject recognition with a
given SNR and spatial frequency. Their capabilities were quite
surprising, in a similar way to how experienced bird-spotters can tell a
bird type when all the casual observer can tell is the colour!

Quite how you relate this to image quality in digital cameras I am not
sure, but I completely agree that both MTF and SNR are important. My
guess is that one should use targets of varying contrast (and ideally sine
wave targets), and measure the SNR within a specified spatial frequency
range (rather than just broadband noise). Perhaps something like an
octave of noise centred on the spatial frequency being tested (I suspect
the 1/3 octave of audio may be too fine).

Nice subject for someone's PhD?

Cheers,
David

While a bar chart target is not a sine wave, which is what MTF is
designed for, a normal bar chart target is a 50-50 square wave. Hence
there is no second harmonic. The first actual overtone that comes
into effect is the third. If there is reasonable anti-aliasing
filtering, then one can use modulation measured from a bar chart to do
MTF calcs. Yeah, a sine wave chart is better, but harder to find,
especially for IR.

.

Relevant Pages

  • Re: Calculation of snr
    ... a normal bar chart target is a 50-50 square wave. ... Indeed, yes, but you can predict the SNR on the retina of a square-wave ...
    (rec.photo.digital)
  • Re: Complex Graph
    ... It's a relatively simple stacked bar chart. ... on whether the county met the goal in that area. ... counties each have 9 areas in which they will be compared with a score ... color-coded by whether or not they met the target percentage. ...
    (microsoft.public.excel.programming)
  • Re: Target line
    ... Jon Peltier has instructions for adding a line to a chart: ... Pat Jennings wrote: ... > How is a target line, say 98%, included in a bar chart? ... > data is less than the target, ...
    (microsoft.public.excel.charting)
  • Re: BB drop outs!!!
    ... The exchange reset my target SNR to 12dB. ... Plusnet arranged to reset it back down to 6dB which worked fine, but a couple of weeks later it wound itself back up to 15dB. ... The impression I got ) was that BT had upgraded the monitoring to take into consideration the error rates on the line as well as simply dropouts. ... It is long term stable at sync speeds around 6300 with SNR varying from 5db at night to 8-9dB during the day. ...
    (uk.telecom.broadband)
  • Re: BB drop outs!!!
    ... The exchange reset my target SNR to 12dB. ... reboot) I've had 103 crc errors downstream and 2 crc errors upstream. ...
    (uk.telecom.broadband)