Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Computing > Digital Photography > How many mp's to match resolut

Reply
Thread Tools

How many mp's to match resolut

 
 
Roger N Clark change user
Guest
Posts: n/a
 
      05-06-2005
David J. Littleboy wrote:
>

wrote in message
> news:(E-Mail Removed)...
>
>>"Roger N. Clark (change username to rnclark)"

wrote:
>>
>>>(E-Mail Removed) wrote:

>>
>>>>Any signal with components less than the Nyquist frequency can be
>>>>reconstructed by using this formula:
>>>>
>>>> sin(pi(t-k)/T)
>>>> x(t) = sum (-inf, inf) x[k] --------------
>>>> pi(t-k)/T
>>>>
>>>>and this is true regardless of the phase of the signal.
>>>>
>>>>(x[k] are the samples, k is the time of each sample, and T is the
>>>>sampling interval.)

>>
>>>This is fine if your do infinite sampling. But imaging sensors
>>>are discrete and finite sampling of a spatial image.

>>
>>Yes, indeed. However, in practice you can get close to the sampling
>>frequency with a window of a fairly small number of samples, say 64 or
>>128. And you can reconstruct the signal from this, regardless of its
>>phase.

Andrew,
How would you reconstruct the image on this page:
http://www.clarkvision.com/imagedetail/sampling1.html
e.g. the "Nyquist sampling" or "1.5x Nyquist sampling?"
> But the only thing you can reconstruct is a sine wave, not a square wave.
> People want images of black and white stripes at frequencies close to the
> Nyquist, and that's completely unreasonable.

But even a sine wave at the Nyquist frequency will be grossly distorted
when sampled at or close to the Nyquist frequency.
> By the way, I checked a portfolio of 4x5 landscape shots in a recent
> Japanese magazine, and fully half of them were shot at f/45. That means that
> the MTF50 from diffraction is a whopping 18 lp/mm. Not a lot of MP there.
> (Well, actually, a lot of MP if you figure 3 pixels per line pair, and 120 x
> 95mm => 33MP or so. But nowhere near the 100MP and over figures people have
> been bandying about.)

David,
You pick one MTF for film, but as you know, that is not the whole
story. Film has higher resolution at lower MTF, and it is that
higher resolution that contributes to image detail. The above
web page illustrates this with a 4x5 image at f/45. At f/45,
the Rayleigh limit is ~9% MTF and detail like grass blades in the
distance are still recorded as lines, and sampling higher than the
Rayleigh limit still brings out a little more detail that
improves the perception of sharpness.
Roger

 
Reply With Quote
 
 
 
 
Marvin
Guest
Posts: n/a
 
      05-07-2005
Roger N Clark change user wrote:
> David J. Littleboy wrote:
>
> wrote in message
>
>>news:(E-Mail Removed)...
>>
>>
>>>"Roger N. Clark (change username to rnclark)"

>
> wrote:
>
>>>>(E-Mail Removed) wrote:
>>>
>>>>>Any signal with components less than the Nyquist frequency can be
>>>>>reconstructed by using this formula:
>>>>>
>>>>> sin(pi(t-k)/T)
>>>>> x(t) = sum (-inf, inf) x[k] --------------
>>>>> pi(t-k)/T
>>>>>
>>>>>and this is true regardless of the phase of the signal.
>>>>>
>>>>>(x[k] are the samples, k is the time of each sample, and T is the
>>>>>sampling interval.)
>>>
>>>>This is fine if your do infinite sampling. But imaging sensors
>>>>are discrete and finite sampling of a spatial image.
>>>
>>>Yes, indeed. However, in practice you can get close to the sampling
>>>frequency with a window of a fairly small number of samples, say 64 or
>>>128. And you can reconstruct the signal from this, regardless of its
>>>phase.

>
> Andrew,
> How would you reconstruct the image on this page:
> http://www.clarkvision.com/imagedetail/sampling1.html
> e.g. the "Nyquist sampling" or "1.5x Nyquist sampling?"
>
>>But the only thing you can reconstruct is a sine wave, not a square wave.
>>People want images of black and white stripes at frequencies close to the
>>Nyquist, and that's completely unreasonable.

>
> But even a sine wave at the Nyquist frequency will be grossly distorted
> when sampled at or close to the Nyquist frequency.
>
>>By the way, I checked a portfolio of 4x5 landscape shots in a recent
>>Japanese magazine, and fully half of them were shot at f/45. That means that
>>the MTF50 from diffraction is a whopping 18 lp/mm. Not a lot of MP there.
>>(Well, actually, a lot of MP if you figure 3 pixels per line pair, and 120 x
>>95mm => 33MP or so. But nowhere near the 100MP and over figures people have
>>been bandying about.)

>
> David,
> You pick one MTF for film, but as you know, that is not the whole
> story. Film has higher resolution at lower MTF, and it is that
> higher resolution that contributes to image detail. The above
> web page illustrates this with a 4x5 image at f/45. At f/45,
> the Rayleigh limit is ~9% MTF and detail like grass blades in the
> distance are still recorded as lines, and sampling higher than the
> Rayleigh limit still brings out a little more detail that
> improves the perception of sharpness.
> Roger
>


You've made a good point. The Rayleigh criteron is based on assumptions that do not apply
to a particular photo as seen by a particular person. The human brain is adept at
integrating information. We see a photo or a scene at high resolution a bit of the time,
as only a part of the eye, the macula, has closely spaced rods and cones. Each of use has
a remarkable imaging system. As with other senses and capabilities, some are better at it
than others. And training can improve matters.

If folks understood this better, there would be less debate in this NG aout optimum
resolution, printing details, etc.
 
Reply With Quote
 
 
 
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Re: How include a large array? Edward A. Falk C Programming 1 04-04-2013 08:07 PM
re.sub(): replace longest match instead of leftmost match? John Gordon Python 13 12-20-2011 02:58 AM
pat-match.lisp or extend-match.lisp in Python? ekzept Python 0 08-10-2007 06:08 PM
$match = true() for empty $match?? Victor XML 2 05-17-2004 10:43 AM
Java regex can't match lengthy match? hiwa Java 0 01-29-2004 10:09 AM



Advertisments