[opendtv] [Fwd: Re: Re: Math of oversampling - a simple comparison]

Arrgh!

I just went thru my Sent folder and there are a few posts on this 
topic that I thought I'd sent to the list on this topic but maybe 
instead just went to Jeroen.  I usually filter mail with {opendtv] 
in the subject into a folder by that name and assume they are from 
opendtv.

So I'm forwarding a couple that should have been to opendtv, 
hopefully not too dated.

- Tom


-------- Original Message --------
Subject: Re: [opendtv] Re: Math of oversampling - a simple comparison
Date: Thu, 28 Apr 2005 20:55:55 -0400
From: Tom Barry <trbarry@xxxxxxxxxxx>
To: Jeroen Stessen <jeroen.stessen@xxxxxxxxxxx>
References: 
<OF51044378.73AEC039-ONC1256FF1.004F2321-C1256FF1.00515866@xxxxxxxxxxx>

Been away all day but I just now added Jeroen's latest two
comparisons to my web page.  The complete sequence is now:


1)  The left&right comparison of the Medium show that started all
this.

      www.trbarry.com/Medium_1080p_compare_qtr_rez.jpg

2)  The left&right comparison of Leno (again the right filtered to
   960x544)

     www.trbarry.com/Leno_1080p_compare_qtr_rez.jpg

3)  The original unfiltered Leno image

     www.trbarry.com/Leno_1080i.jpg

4)  Jeroen's Leno image creating by scaling to 544p and then back
up to 1088p. Possibly softer but with fewer artifacts.

     www.trbarry.com/Leno_1080_544_1080p.jpg

5)  Jeroen's Leno image creating by scaling all the way down to
480p and then back up to 1088p. Definitely softer but again with
fewer artifacts.

        www.trbarry.com/Leno_1080_480_1080p.jpg

6) Jeroen's Leno 1080->576->1080

        www.trbarry.com/Leno_1080_576_1080.jpg

7) Jeroen's Leno 1080->720->1080

        www.trbarry.com/Leno_1080_720_1080.jpg


I don't know if we would all agree about how much resolution is
really needed here.  But it does look like some sort of law of
diminishing returns sets in above about 544p where there are only
marginal improvements for increasing resolution.

I'm going to stick with my earlier opinion that once we are into
the toe/tail of the MTF curve it is hardly worth the effort.

Jeroen - Do you by any chance have a very high resolution 1080p
image you could post to show an example where it really DOES make
a large difference in, say, 1080->544->1080?  This one wouldn't
even have to start from HD but it would be nice to really see the
difference in one where it matters.

- Tom


Jeroen Stessen wrote:

> Hello, 
> 
> At Craig's request: 
> 
>>If Tom and Jeroen have time to play, I would like to see a down/up 
>>sample pass through 1024 x 576P. 
> 
> 
> Here are the results: 
> (2 files attached). 
> file <Leno_1080_576_1080p.jpg> was created 1920x1088 -> 1024x576p 
> -> 1920x1088. It is still not quite as sharp as we had hoped ! 
> 
> Then I have also made file <Leno_1080_720_1080p.jpg> : 
> 1920x1088 -> 1280x720p -> 1920x1088. Now that is finally almost 
> as sharp as the original. It appears that we need 1280x720 pixels, 
> that 1024x576 is just too little for this image. 
> 
> (But I am watching this on a 20" CRT in 1600x1200 resolution 
> from 1 foot distance, so that is not fair even for HDTV !) 
> 
> 
>>My educated guess is that it will be nearly impossible to see 
>>the difference from the original.
> 
> 
> Well, you seem to have guessed wrong, but why ? 
> 
> This morning I have analysed my down- and up-sampling filters, 
> and I have determined that the cascade of the two filters has a 
> -3 dB point at 73% of the Nyquist limit of the down-sampled image. 
> So even if you apply this filter at the original resolution, it 
> is still a not neglegible low-pass filter. Unfortunately that is 
> the price to pay for low-aliasing scaling, I can't do much better. 
> Maybe that is why I need a 25% higher sample rate to get the sort 
> of sharpness that you educatedly guessed. 
> 
> Regards, 
> -- Jeroen 
> 
> +-------------------------------+------------------------------------------+
> | From:     Jeroen H. Stessen   | E-mail:  Jeroen.Stessen@xxxxxxxxxxx |
> | Building: SFJ-5.22 Eindhoven  | Deptmt.: Philips Applied Technologies |
> | Phone:    ++31.40.2732739     | Visiting & mail address: Glaslaan 2 |
> | Mobile:   ++31.6.44680021     | NL 5616 LW Eindhoven, the Netherlands |
> | Pager:    ++31.6.65133818     | Website: http://www.apptech.philips.com/ 
> |
> +-------------------------------+------------------------------------------+
> 
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 


 
 
----------------------------------------------------------------------
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: