Convolution 3d?

[ookzDVD]

@forum,

Sorry, I'm new to this filter,
I'm a little bit confuse at this time.

Is that filter will be useful for regular film (from DVD),
not anime and not TV-Cap ?

[bb]

It will be useful for all types of sources.
For DVD it can increase compressibility.
For anime, it can clean up the picture.
For TV captures it can greatly denoise the video.

bb

[ookzDVD]

@bb,

Thank you,
I'm a little bit clear now ;)

I hope more people will share the parameter setting here,
especialy for DVD source.

Thank you.

[SansGrip]

I hope more people will share the parameter setting here,
especialy for DVD source.

For DVD I've seen this recommended:

Convolution3D(0, 4, 4, 4, 4, 2.8, 0)

But the best thing to do is try it for yourself. Since DVD is (usually) already pretty well denoised, you don't wanna go nuts with it unless you're trying to fit Saving Private Ryan on 1 CD.

Remember that the first number is either 0 or 1, with the latter being a slightly more aggressive matrix. The next 2 numbers control spatial smoothing, which is what gives the really noticible softening to an image, and the 2 after that temporal smoothing, which can work miracles on noisy sources like VHS captures. I have no clue what the next number does ;)

You can usually apply significantly more aggressive smoothing to the chroma, since it's less noticible to the naked eye.

Play with the values and see what you come up with. What I usually do is use StackVertical to see the original material and the softened material at the same time, or Subtract to see the difference.

[iago]

Convolution3D(0, 4, 4, 4, 4, 2.8, 0)

... the first number is either 0 or 1, with the latter being a slightly more aggressive matrix. The next 2 numbers control spatial smoothing, which is what gives the really noticible softening to an image, and the 2 after that temporal smoothing ... I have no clue what the next number does ;) (SansGrip)

The next number represents "temporal influence" ;). And it's not a good idea to go above 3 for normal film - DVD sources imho. You can gain some more speed by decreasing it a bit (ie. to 2.8 or 2.6) and preserve a bit more details.

regards,
iago

[wotef]

what's your take on temporal influence? how comes 2.8 has become the sweet spot?

i always need the luma temporal cleaning to kick in for my vhs clean ups, so i set influence to 10, but i'm a bit dense and i'm not perfectly clear on the concept described in the readme

Temporal influence :
I first build a limit = Temporal Luma Threshold * Temporal influence
For each 2 pixel computed (due to MMX, 2 pixel at the same time), I first check this :
if
(Abs (Y0 - Y0[Previous frame]) +
Abs (Y0 - Y0[Next frame]) +
Abs (Y1 - Y1[Previous frame]) +
Abs (Y1 - Y1[Next frame])) > limit
then
do Spacial work (only 3*3 matrix)
Else
do Spacial and Temporal work (3*3*3 matrix)

[ookzDVD]

@SansGrip @ iago,
Thank you for the hint.

PS. I've found this from the readme.txt : Convolution3d (0, 8, 8, 8, 8, 3, 0)

[JohnMK]

Isn't that a bit aggressive for your purpose?

[JohnMK]

quick question: to do compressibility check, do I set single-pass quality-based to 100%?

I want to compare compressibility of standard DVD vs. convoluted 3d DVD.

[ookzDVD]

@JohnMK,

Simply run the 2-pass 1st pass,
for each test you have to use Koepi's stats reader to see the 1st pass size, just remember all the 1st pass parameter must be the same
for each test.

The one smaller is the one more compressable ;)

Goodluck ! ;)

[JohnMK]

But is my method equally as workable?

[ronnylov]

Originally posted by JohnMK
But is my method equally as workable?

What encoder are you using? You wrote that you would like to compare DVD with and without concolution3D.

DVD is MPEG2 so I guess you are talking 1-pass TMPGEnc with CQ100% to create a DVD compliant file? This can be used as compressibility check. The smaller resulting filesize - the more compressibility! But what to do next? If the filesize is 10 % smaller does that mean I can use 10 % less average bitrate in a 2-pass encoding and get the same quality in my DVD encodings?

[JohnMK]

Encoder: divx 5.02, gordianknot -> virtualdub.

[Aktan]

It workable, just like ronnylov said, the smaller the file size, the more compressible it is. But doing 2-pass 1st-pass requires a lot less space.