Posted by Smarty on 11/03/07 00:10

Martin,

My personal approach is to use VideoReDoPlus to select and capture the frame
to the clipboard, then to paste the clipboard directly into Photoshop, and
then save the image as a TIFF file, nominally about 6 MBytes in size for a
single frame. There is no JPEG whatsoever in my method, and all detail of
the frame is preserved.

"Jaggies" along a diagonal path are not only expected but conceptually
necessary and appropriate. The edge of a diagonal transition ultimately
***MUST*** cut pixel boundaries, and either be:

1. represented by a staircase ("jaggies")

or

2. be smoothed with an "anti-aliasing" algorithm which turns a crisp and
abrupt brightness/color transition into a less sharp blur / smear.

If one were asked to draw a diagonal line on engineering "quadrille" or
graph paper, one would be faced with precisely the same two alternatives,
since a rectilinear grid of pixels cannot exactly fall at each and every
diagonal point.

Low pass spatial filtering of the scene by reducing sharpness / blurring
diminishes this problem, reducing the abrupt transition, seen in the time
domain as a very fast rise time pulse edge, or in the frequency domain as
very high frequency energy. A black and white checkerboard stripped of its'
rich high frequency components eventually becomes a gray and white polka-dot
array, where square waves have been stripped down to their fundamental sine
wave spatial frequency with smooth and continuous sine waves replacing sharp
and abrupt square waves. Canon and other HD camera designers chose the
default "sharpening" to preserve this detail, limited primarily by the mpeg
encoder's processing bandwidth. Creating too much detail for the mpeg
encoder to handle is a sure recipe for errors in encoding when the camera or
subject undergo rapid changes / pans / zoom / etc.

Smarty


"Martin Heffels" <goofie@flikken.net> wrote in message
news:t4smi39nm9g3k1lq0n6ra4euoq90dvkqd5@4ax.com...
> On Fri, 02 Nov 2007 14:43:18 GMT, "Smarty" <nobody@nobody.com> wrote:
>
>>being pretty typical of what I would expect for the HV20 with additional
>>JPEG processing applied.
>
> I notice the same thing. Looking at the yellow lines (on the black), you
> can see similar artifacts in the rest of the picture.
> How do you look at all those details? Do you export a frame from your NLE?
> And in what format? Did you export to TIFF or Targa, and view them?
>
> -m-
> --
> Official website "Jonah's Quid" http://www.jonahsquids.co.uk

[Back to original message]