Re: compensate for 'pixel shimmer' on LCD screens somehow?

In article <1191257078.856720.229420@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>, jkn
<jkn_gg@xxxxxxxxxxxxxxx> wrote:

Thanks for the info, and useful links - but It's not an interlacing
artifact I'm talking about. I know what that looks like, and
(typically) I'm talking about horizontal movement (eg. of a face)
anyway.

Another way of describing the artifact I'm seeing is to look at a
person's shiny hair under studio lighting. If you pay attention to the
whole mass of the hair, then the brightness of the *whole mass*
changes as their head changes from still, to moving, to still.

In a hand-waving kinda way, It feels like the DCT is truncating the
sharp edged of the hairs at (say) a 'bright' level when still, but a
'dim' level when moving, or vice versa. But I'm not saying it *is* the
DCT; I'm aware that there are many levels of processing going on.

I see I shall have to experiment quite a bit before buying a new TV!

I believe that what you are describing is an inherent problem in
current compression algorithms, rather more than a bandwidth issue -
although the two are obviously related. Typically, when converting film
or analogue video to MPEG, the algorithm works on each frame based on
the degree to which it has changed from the previous frame. If working
on footage of, say, a long putt rolling across a green where the first
part of the sequence shows the ball moving faster and a degree of
motion blur in the green background, a harsh algorithm will want to
select the ball as the only thing which moves, although the grass is
also moving because the camera is panning across it. Hence, a harsh
algorithm will initially try to represent the blurred background with
static pixels and will only detect change in the background as the golf
ball slows down, making the transition look smeary, blocky and
unrealistic There are other factors which come into play, including the
number of colours recorded in the footage and the number of colours a
digital TV can display. The green is comprised of dozens, perhaps
hundreds of colours, mostly different shades of green but many others
too. Afaik, there are no LCD or plasma screen which are capable of
showing them all, even in the unlikely event that they are included in
the original footage. The increasing use of HD digital cameras will at
least bring the resulting TV picture much closer to the original
footage (depending upon how many colours the selected compression
algorith decides to drop out) and so the issues you are complaining
about will gradually reduce. It is already possible to see clear
improvement in the definition of a football pitch viewed in SD when the
original footage was shot with a HD digital camera.

Stan
.

Relevant Pages

  • Distance normalized TSP algorithm
    ... was the the algorithm required the distance between each of the nodes, ... Then T_2 considers moving to all other nodes available, ... Notice that with the distance normalized algorithm you just have cost ... single traveler moving along the path of T_1, ...
    (comp.lang.java.programmer)
  • vb.net code for advanced data structures
    ... Given is a dynamic set S of (moving, vanishing and arising) points ... In general almost all of the current members of S are moving all the ... It would be fine to have an algorithm which also can handle ... correct voronoi diagram for moved objects without a complete new ...
    (microsoft.public.dotnet.languages.vb)
  • Re: a chessboard type problem
    ... a real number (pos or neg). ... You want to navigate from one side to the ... other by either moving one space ahead or one space ahead and one ... algorithm would be great. ...
    (comp.theory)
  • sequnce - is there any better algorithm ?
    ... for simplicity let's assume that we have 4 people. ... I am moving toward beginning: now second person says that two people from his left side should be before him. ... As we know that 4th is before 3rd we simply shift 2nd person two positions left: ... But it's Oalgorithm. ...
    (comp.programming)
  • Re: Determine if a moving object will pass within range of point
    ... > Need an algorithm or series of algorithms to determine if a moving ... > object will pass within set range of another point. ... to a flat-earth approximation good for short distances)? ...
    (sci.math)