[opendtv] Re: [Fwd: Re: Re: Math of oversampling - corrected links

  • From: Jeroen Stessen <jeroen.stessen@xxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Mon, 2 May 2005 11:51:45 +0200


Dale Kelly wrote: 
> The 1080P image appears to contain significant "interlace" type 

This may well be an artefact from conversion from .bmp to .jpeg . 

The original picture was a .jpeg, with properties that I can not seem 
to extract. I've then cropped it to 1920x1080 and saved it as .jpeg 
again (with a mild compression factor, as we all have broadband internet). 

Conversion to .jpeg always gives considerable loss, even when it is 
"lossless", because the YCbCr components make less efficient use of the 
3x 8-bits code space than the original RGB components did. 
So we can expect quantisation errors from RGB->YCbCr->RGB conversions. 
But that does not explain what we are seeing here. 

I know that I have selected sub-sampling ("4:2:0" in video terms) of 
the CbCr chroma components this time. This is certainly suspicious. 
For the resolution-reduced images this will be less critical. 

The visibility of the artefacts from chroma down- and up-sampling 
depends on the respective algorithms for down- and up-sampling. 
It may be that different image viewers use slightly different algorithms 
for up-conversion back to 4:4:4, so that we are not all seeing the same 
artefacts ! I know that our pfspd viewers deliberately use pixel 
repetition "to see what's in the file", where a filter would be advised. 

If you say "interlace" type artifacts, it could also be that you are 
referring to sampling artefacts in general. Those slanted white lines 
have been made by somebody applying a rotation algorithm to an 
originally rectangular image. This requires a sample-rate conversion 
operation, and we don't know which kind of filter he/she has used. 
It may have been the standard bicubic interpolator, which may give 
some more sampling artefacts than my own choice of polyphase filter. 
These artefacts will disappear by subsequent down-sampling with a 
better filter, which then dominates because of its lower bandwidth. 

Remember that pure progressive images can have as many sampling 
artefacts as interlaced images. (Interlacing is over-sampling, I say.) 

-- 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 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: