Software filtering question..

[rbatty11]

Is this a true statement: When using software filters, original source must always be re-encoded --> re-encoding leads to loss of quality.

If true, I'm trying to understand why so many opt for software filters.. Is it b/c the 'enhancement' of the image when using software filters is worth the loss in quality?? (sorry, if I'm using incorrect terms).

[gadgetguy]

The statement is true, but the degree of loss can be negligable and not visible when watching the video at normal speeds on normal TVs. It depends on what filters you use and what goals you're trying to accomplish.

Only you can decide if the efforts are worth it.

[mats.hogberg]

Only time I use filters is when I have to reencode anyway.

/Mats

[jimmalenko]

I'll split this up:

When using software filters, original source must always be re-encoded

True.

re-encoding leads to loss of quality.

That's not a hard and fast rule in my book. If your source is of questionable quality, then it's normally more apparent. If you've got an excellent quality source, then it is far less apparent. Also, there is the case of an umcompressed source, you run some filters on it, and output uncompressed. I fail to see any possibility at all that this re-encode degrades quality.

If true, I'm trying to understand why so many opt for software filters.. Is it b/c the 'enhancement' of the image when using software filters is worth the loss in quality?? (sorry, if I'm using incorrect terms).

As above, filters can only "tweak" - they're not miracle workers. I firmly believe that the right filters at the right time can improve the visual quality. Otherwise, as you say, why would people filter if they're not enhancing the video at all ?

[FulciLives]

Filters were designed originally to work with a high quality video source such as HuffyUV or PICVideo MJPEG or DV where you capture in that high quality format then need to convert (compress) down to something else (such as MPEG-2 DVD spec).

So in the encoding process why NOT add filters ... if they are needed?

- John "FulciLives" Coleman

[vhelp]

Because I am in the middle of an exhaustive RGB and YUV image
research and discovery quest, I feel I have my own set of
opinions to share on this

.
.

rbatty11 wrote:

re-encoding leads to loss of quality.

Well.., IMHO, this is true. re-encoding always leads to some
level of quality loss.

But, maybe "loss" is not the right word to use here

(at least at this time in my dealings) I would rather replace the
word "loss" with "detail", but..

The reason why I feel this way, is dependant upon the type of
*method* used during the pixel manipulation. And there are many
methods of pixel manipulations. Though most are one variation or
another of "averaging" pixels on a number of images. So lets use
this "average" as a theory and take a short tour of its use..

The moment you take away from the original source, is the first acts
of quality loss or detail.

Ok. For instance, say you have a pixel who's R/G/B values are a certain
set of numbers.. ie, 255 / 255 / 255

( or, to get the true color of a certain pixel, you would perform the
following component calculation: 255*255*255 = 16,581,375 )

and then you apply an average on this pixel against another nearby pixel,
who's R/G/B values are a certain set of numbers.. ie, 100 / 90 / 16

( or, to get the true color of a certain pixel, you would perform the
following component calculation: 100*90*16 = 144,000 )

When speaking of "averaging" (as an example) the user would take the
R/G/B components each, (that's why I separated them w/ an "/")
and perform the averaging.

In the example above:

255/255/255 and 100/90/16 (or 16581375 and 144000) average computes
would be:

(R=255 + R=100)/2 = R=177
(G=255 + G=090)/2 = G=172
(B=255 + B=016)/2 = B=135

Thus, the Averagel_New_Pixel_TrueColor_Value = 4,109,940

( or, broken down into their respective components: 177/172/135 )

The point of this excercise (or theory) is to show how *ONE* pixel
changed would result in a perspective of quality loss or detail.

Thus, once you change one pixel's value (not mentioned, hundreds of'em)
you change the original sources detail, or consider there to be some
level of quality loss.. depending upon once's perspective of "loss"

fwiw noting here, the above could just have easily been used inside an
YUV color space.

However, there (IMHO) another side to this quality loss theory.
But, I feel it crosses the boarder of this topics subject.. at this
time.

-vhelp 3759

[rbatty11]

OK, I have a better understanding now.. Thanks.

So what is consensus here on which method is "superior" for transferring VHS/LD to DVD?? Using all hardware (full frame TBC, color corrector, detailer, DVD recorder, etc.) or capturing to pc and using software filters?? Assumption: Source material is NOT high quality.

I'm doing xfrs of old home videos for friends using a JVC DVHS, full frame TBC and a JVC DVD recorder, then editing on pc, authoring.. I just want to make sure I'm not "cheating" them on quality.