Try Quenc for DV encode

[insertdisk]

Just to let everyone know - there is a free tool called Quenc (other mpeg encoders) and it does a real fine job on DV footage. Actually it is an encoder, I think it does a real good job anyway.

If anyone else trys it out, please post what you think about it.

[bb]

Note that QuEnc needs an AviSynth script as input, and that the script must deliver YV12 frames (so if in doubt use ConvertToYV12 at the end of your script). Make sure you didn't miss the "Advanced Options" button ;)

I did a small test with a progressive 16:9 DV type-1 video using QuEnc and CCE 2.66.01.07 at similar settings:

QuEnc1: MPEG-2 4000 kbps, Use VBR, High Quality, DC Precision 9, Scene Detection, 8500 kbps max. Bitrate, 16:9 Aspect Ratio

QuEnc2: MPEG-2 4000 kbps, Use VBR = OFF, High Quality, Use Trellis Quant, DC Precision 9, Scene Detection, 8500 kbps max. Bitrate, 16:9 Aspect Ratio

QuEnc3: MPEG-2 4000 kbps, Use VBR = OFF, High Quality, DC Precision 9, Scene Detection, 8500 kbps max. Bitrate, 16:9 Aspect Ratio

CCE: MPEG-2 4000 kbps, CBR, DVD Video compliant, Upper field first = OFF, filters turned off (Detailed setting), Quantizer characteristics 16, DC precision 9, Block scanning order = Zigzag, Progressive frame flag

Visual test using
cce=MPEG2Source("C:\Filme\DV\CCE.d2v")
quenc1=MPEG2Source("C:\Filme\DV\QuEnc1.d2v")
quenc2=MPEG2Source("C:\Filme\DV\QuEnc2.d2v")
quenc3=MPEG2Source("C:\Filme\DV\QuEnc3.d2v")

cce=cce.SimpleResize(1024, 576).Crop(512,100,512,300).Subtitle("CCE")
quenc1=quenc1.SimpleResize(1024, 576).Crop(512,100,512,300).Subtitle("QuEnc VBR")
quenc2=quenc2.SimpleResize(1024, 576).Crop(512,100,512,300).Subtitle("QuEnc CBR Trellis")
quenc3=quenc3.SimpleResize(1024, 576).Crop(512,100,512,300).Subtitle("QuEnc CBR")

StackVertical(StackHorizontal(cce, quenc1), StackHorizontal(quenc2, quenc3))

Results:
--------
QuEnc1:
filesize = 23.721 KB
speed ca. 15 fps
quality: bad

QEnc2:
filesize = 23.547 KB
speed ca. 7 fps
quality: good

QEnc3:
filesize = 23.550 KB
speed ca. 15 fps
quality: good

CCE:
filesize = 23.418 KB
speed ca. 23 fps
quality: best

QuEnc seems to have problems with the VBR mode. The first frame was extremely blocky, the following frames were normal, but less in quality compared to the CBR encodes. The Trellis quantisation doubled the processing time, but the quality improvement was only marginal.

CCE was very close to the original (including the noise) in I- and P-frames; B-frames were smoothed somewhat. QuEnc smoothed the noise out in all frames and showed more blocks.

Not that QuEnc delivered bad quality, but without any filtering enabled the CCE encode came out better. If you add filtering to your AVS, the result may be completely different, I don't know. As I said above, it was just a small, quick test, a first impression.

bb

[insertdisk]

I did a few tests but didn't document things quite as well as you bb:p

Anyway my cam shoots interlaced and so I encode interlaced as well. I noticed the best results with all quality improvement options checked DC precision set to 10 and scene detection off. 2 pass VBR 6000 bitrate.

I compared against procoder on the TV set and I thought Quenc looked a bit better ( not as much washout or loss of detail in lighter areas of the video ) - specifically when the sun shines on my daughter's face there seems to be alot of detail loss using procoder.

My clips were fed using convultion 3d as filter - I will post script later.

I just wanted people to be aware of the possibilities.

In my tests Procoder is probably a bit faster at the encode.

Thank you for the test results and quick response - I will definitely take a closer look before I switch to this encoder for archiving all of my DV footage to DVD.

Here is the AVS that I use - and I also looked and the file I am converting was Hi8 that had been imported through my DV cam.

******
LoadPlugin("C:\Pro~\AviSynth 2.5\plugins\ReInterpolate411.dll")
LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\convolution3d.dll")
AVISource("D:\video\Han2-29-04.04-02-29_14-44.00.avi")
Reinterpolate411()
separatefields()
vid_e=selecteven()
vid_o=selectodd()
vid_e=convolution3d(vid_e,0,32,128,16,64,10,0)
vid_o=convolution3d(vid_e,0,32,128,16,64,10,0)
interleave(vid_e,vid_o)
weave()
ConvertToYV12()
********