[opendtv] Re: Panasonic's AG-HVX200

  • From: Tom Barry <trbarry@xxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Thu, 28 Apr 2005 09:49:07 -0400

Alan Roberts wrote:

> The mtf (lower case) is the "modulation transfer factor". It is, at any spot
> frequency (or aperture, or focal length, of position with the frame) the
> gain of the device. So, if you put a sinusoidal signal into the device
> (lens, display, camera, whatever) at a particular frequency, the mtf is the
> multiplier that it applies to that frequency.
> 

Alan -

Thanks, I've seen the Canon article.

But mtf must be more than just a multiplier factor.  If I take a 
sinusoidal signal and multiply it by .5 then the amplitude is only 
half as much but no information has yet been lost, barring 
quantization.  I could multiply it by 2 later and get it back.  It 
is only when errors (random info) are introduced that information 
goes away.  But I'm not sure how this figures in gain or mtf and 
I'm not sure how that  part is supposed to be measured.

I realize there are many sources of errors, A/D noise, 
quantization noise, various optical, electronic, and digital 
filters etc. and each contributes in complex ways to lowered mtf 
(either case).  But I still can't quite visualize what is being 
measured when "contrast" is lowered.

What am I missing here?

- Tom


> The MTF (upper case) is the "Modulation Transfer Function" and is the
> distribution of mtf over the whole frequency/aperture/focal length/position
> for the device. It is therefore a multi-dimensional curve which
> characterises the response of it to sinusoids at all possible settings and
> positions of it.
> 
> So, for a lens, you'll see a 1-dimensional curve showing what's effectively
> the frequency response (at one aperture, focal length, centre of frame). And
> you'll get another showing how that varies with position in frame (at one
> frequency, aperture, focal length). And another showing how that varies with
> focal length (at one frequency, aperture, position). Etc.
> 
> You might like to look at this article for Canon's take on how their HD
> lenses compare with their SD lenses, but remember that every part in the
> transmission chain, from light in to light out, has a set of MTF curves, and
> it's the convolution of all of these that gives the overall performance of
> the system..
> 
> http://broadcastengineering.com/mag/broadcasting_hdtv_lens_design/index.html
> 
> ----- Original Message ----- 
> From: "Tom Barry" <trbarry@xxxxxxxxxxx>
> To: <opendtv@xxxxxxxxxxxxx>
> Sent: Thursday, April 28, 2005 1:46 PM
> Subject: [opendtv] Re: Panasonic's AG-HVX200
> 
> 
> 
>>I have never really understood the MTF curve.  It appears to be a
>>graph of how much contrast is available at the various
>>frequencies, starting out high and declining so there is little or
>>eventually no contrast past a certain high spatial frequency limit.
>>
>>And the perceived sharpness is considered to be the square of the
>>area under this curve.  This makes sense since the area under the
>>one dimensional curve would just be the average contrast over the
>>  spectrum and an image is 2 dimensional so (I think?) the average
>>would decline as the square of the area under the curve.
>>
>>But what exactly is this "contrast" that is being measured?  Can
>>anyone explain this?
>>
>>- Tom
>>
>>
>>Craig Birkmaier wrote:
>>
>>
>>>At 10:16 AM -0400 4/27/05, Mark Schubin wrote:
>>>
>>>
>>>>>Now a question. Is not sharpness directly related to the ability to
>>>>
>>>>>perceive detail?
>>>>>
>>>>>
>>>>
>>>>No.
>>>>
>>>>Sharpness is proportional to the square of the area under an MTF curve.
>>>>An MTF curve plots contrast ratio against detail.
>>>>
>>>>In my PBS presentation, I extracted two sections from a contrast
>>>>sensitivity grating.  One, taken from the low-detail high-contrast
>>>>section clearly appears to be sharp.  The other, taken from a
>>>>moderate-detail moderate-contrast section does not.  I didn't choose
>>>>high-detail low-contrast for the second, because then the audience
>>>>wouldn't have been able to distinguish it from a grey block.
>>>
>>>
>>>Thanks again Mark. This has always been a bit confusing to me. I
>>>suspect that the situation has been amplified a bit as my eyes have
>>>aged and I have been forced to use reading glasses and seek brighter
>>>light in order to read "the fine print."
>>>
>>>What it seems to boil down to is the contrast ratio of the detail
>>>that one wants to perceive. In your grating examples the contrast
>>>ratio keeps getting lower as the the level of detail increases.
>>>
>>>While you properly point out that there are many contributing
>>>factors, I  believe that one of the major issues is the need to
>>>properly filter images to obey the laws of physics and sampling
>>>theory. For video imagery we cannot allow unfiltered transitions from
>>>black to white, as is common with the non-Nyquist  limited samples
>>>you are looking at as you read this message. So while the text on
>>>this page may have a black sample next to a white sample, in a video
>>>image we need to ramp from black to white - through grey - or to
>>>limit the contrast of the transition.
>>>
>>>At the lower frequencies of the grating there is good contrast
>>>because the peak intensity of the samples reaches full white and full
>>>black. There are grey samples in between, but the frequency of the
>>>detail is low enough to be represented faithfully. As the frequency
>>>of the detail increases, however, we reach a point where the
>>>intensity of the stimulus can no longer be represented with full
>>>black or white samples. Filtering becomes a factor causing the
>>>samples to be limited in contrast as the frequency increases, until
>>>we reach the point where the grating becomes a grey blob.  Obviously
>>>filtering is not the only contributing factor. Most cameras employ
>>>optical filtering, and the lens itself limits what the sensor can
>>>"see." Hence, the MTF curve does a good job of predicting the
>>>relationship of the original stimulus to the intensity of the
>>>stimulus that we see in the sampled image. More area under the curve
>>>is another way of saying that a system can pass full contrast
>>>transitions at higher frequencies, thus enhancing our ability to
>>>perceive higher levels of detail.
>>>
>>>And then there is the reality that the display ALSO has an MTF curve,
>>>and that it can impact the ability to perceive detail based on the
>>>contrast, brightness and sharpness controls. Joe Kane makes his
>>>living in this space, teaching people how to control ambient light
>>>and adjust a display so as to maximize perceived contrast and detail.
>>>
>>>Does all of this make sense Mark?
>>>
>>>Regards
>>>Craig
>>>
>>>
>>>----------------------------------------------------------------------
>>>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.
> 
> 
> 
>  
>  
> ----------------------------------------------------------------------
> 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: