[opendtv] Re: Image quality
- From: Tom Barry <trbarry@xxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Sun, 13 Nov 2005 16:24:13 -0500
Mark Schubin wrote:
> You can add bits at the initial imager A/D to help with this. Many
> cameras use 14-bit A/D converters, and there are valuable reasons why
> they do so. But, if you're dealing with half of those levels, you're
> still 6 dB down. Half of what's left is another 6 dB. Half of that is
> another 6 dB, and so on. An MTF of 6.25% gives you 24 dB less SNR to
> deal with -- and that was before the lens reduction of MTF. If that's
> enough for the processing you want to do, great! If not, you may be
> creating a white or black dot that weren't there in the first place.
I've heard adding extra bits to an A/D can sometimes just give you extra
bits of noise if you aren't careful. However let's (falsely?) assume
that doesn't happen because someone spends enough money on it. Then an
MTF of 6.25% on a 14 bit converter might lose you 4 bits and leave you
10. Dunno if that is good enough or not but it doesn't seem to
automatically rule out doing anything useful.
Anyway, I guess it points out that better electronics might or might not
compensate somewhat for lesser optics. What you said was pretty scary
about the cost of some of those lenses. But additional silicon seems to
be increasingly free along the progression of Moores Law these days.
- Tom
> Tom Barry wrote:
>
>
>>I get confused in going back and forth between computer and TV terms so
>>I may be saying something stupid here. But in my own playing with
>>various sharpening algorithms I always eventually run up against the
>>point where sharpening or whatever starts to just amplify quantization
>>or other noise. That is, it works well for high bit precise source but
>>not very well for whatever 8-bit previously compressed source I might be
>>playing with. I realize deconvolution is much more complex but the same
>>principles likely apply.
>>
>>
>
> Agreed.
>
>
>>It appears you equate limited dynamic range with limited MTF at higher
>>spatial frequencies. But that limited MTF can come from many sources
>>and one of those is probably adjacent pixel averaging caused by your
>>optical limitations. And it is that last factor we would try to correct
>>if we got the signal in the camera before too much other noise was
>>introduced.
>>
>>
>
> It doesn't matter what reduces the level. Suppose we have an eight-bit
> system in which peak white is level 255 and deepest black is level
> zero. That's 48 dB of SNR plus any improvement associated with the form
> of video. Let's ignore the improvement for the moment. If you need to
> recover a white and a black line from half the levels, you've got 42 dB
> of SNR to deal with. If a quarter of the levels, 36 dB.
>
> You can add bits at the initial imager A/D to help with this. Many
> cameras use 14-bit A/D converters, and there are valuable reasons why
> they do so. But, if you're dealing with half of those levels, you're
> still 6 dB down. Half of what's left is another 6 dB. Half of that is
> another 6 dB, and so on. An MTF of 6.25% gives you 24 dB less SNR to
> deal with -- and that was before the lens reduction of MTF. If that's
> enough for the processing you want to do, great! If not, you may be
> creating a white or black dot that weren't there in the first place.
>
> TTFN,
> Mark
>
>
>
> ----------------------------------------------------------------------
> You can UNSUBSCRIBE from the OpenDTV list in two ways:
>
> - Using the UNSUBSCRIBE command in your user configuration settings at
> FreeLists.org
>
> - By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
> unsubscribe in the subject line.
>
>
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
