On 10 Apr 2003 01:41:20 -0700, gzooflup@my-deja.com (Jerome Marot) wrote:

>d_ruether@hotmail.com (Neuman - Ruether) wrote in message news:

<3e972ea7.4379558@newsstand.cit.cornell.edu>...

>[...]

>> Thanks to all for your comments - I have copied them all,

>> and will try to get through them as I can. BTW, as a "test",

>> I used a video I produced a while back that has a very

>> difficult scene in it for compression: a garden full of

>> small flowers shot with a VX2000 turned, then moved forward

>> through a gate in a rock wall, then turned back to look

>> again into the garden.

>There is a problem with this approach in that a scene with lots of

>details (flowers) is already difficult to compress to DV. The

>resulting data causes problems for mpeg, because the DV artifacts add

>"noise" to the picture and noise compresses badly.

Yes - but these conditions do appear in DV videos shot

with sharp-imaging cameras, and the original is acceptable.

I do not want to soften these scenes (or do much else)

to optimize MPEG2 results. Without this, there are still

clear differences in how different encoders handle

this situation, ranging from unacceptable to acceptably

close to the original... It is therefore a good test

for my purposes...

>There are two different problems with mpeg compression: the first one

>is to decide what part of the picture data one will throw away. This

>is not too difficult and normally, the operator can add some input

>here. For example, using tmpeg, if you have noisy input data, you can

>add massive noise reduction as preprocessing. This will basicall throw

>the noise away and make compression easier, but also make the image

>softer. Your test shows that some encoder take this approach all the

>time, which works very well for many people... because the average TV

>is not very sharp to start with.

Yes. The CinemaCraft Basic appeared to do this, with

unacceptably soft results (I do have a very sharp TV...;-);

the others softened the image *very* slightly, as I would

expect from going to the DVD medium...

>The second problem is to find out what parts are similar between

>following pictures (what is called "motion vectors estimation"). This

>is really hard and is the part where good encoders can really gain a

>lot.

>[...]

Yes. The MyDVD and Ulead run not-RT did fairly well,

the RT use of the MyDVD failed on this scene, and the

Panasonic recorder handled it very well (I may get one

[RT, and good results, too - with the ability to store

encoded files for multiple-DVD writing later...]). I

would expect good VBR multi-pass software to be excellent,

too. Part of what I want is encoding speed, so the

Panasonic hardware may be the best solution for me...

>> work (also BTW, we do not EVER load QT on our computers ;-),

>Many video or DVD editing software loads it behind your back.

NOT if I can do something about it! ;-) Like, dump the

load when QT is discovered...;-)

>[...]

>>I'm looking for a

>> "straight-forward" encoder that works well with little

>> modification of the original material, works with fairly

>> short render times, and works with ease of use.

>Tmpeg fills the bill, except the "short render times" part, especially

>if you have it turn noise reduction on by default. CCE probably fills

>the complete bill, but is costly.

Yes. The cheap Basic form may not be adequate...

TMPGenc has always been impressive (when it worked

for us...), but, depending on encoder time, may not

serve us, since other, faster solutions appear to be

adequate...

>> While I

>> know that better results can be had with more effort and

>> time, "good enough" is good enough for what I want...

>It is very little effort to look at you data and decide upon a set of

>strategies adapted to the most common encoding difficulties: "noise"

>and "lots of motion".

Not if universal - I value good sharpness over the

occasional sub-standard area - and some encoders appear

to handle these fairly well... If VBR can "save" the best

footage from degradation, and take care of the worst,

that is worthwhile, if encoding does not take too long...

>> Another BTW: I recently tried a Panasonic stand-alone

>> recorder, and liked its encoding more than that of any of

>> the successfully-tried software - but it failed to write

>> disks properly after about 4 minutes (another will

>> betried...).

>I have see results from a stand-alone Panasonic DVD writer (in "2

>hours per DVD" mode, PAL, cable TV source) and was not impressed.

Try it in "XP" mode - it records only one hour on the

DVD (these disks do have only 1/2 the capacity of commercial

DVDs, as you know...), but at very high quality. I expect

to get no more than 1 hour per DVD disk with any encoder...

In the average-difficulty areas, it was as good as any

of the encoders I tried; in the most difficult section,

it was better than any software I had working...