Often, I have avchd 1080i PAL files that I want to convert to 720p NTSC. Generally, I have good results, but occasionally have source files that don't seem to transcode/convert well. Getting from PAL to NTSC is the easy part. Getting a good transcode of the source avchd file to 720p is the problem I having this particular time. I am using DGAVCindex, MeGUI, and a fairly simple avisynth script. The result is okay, but I am seeing quite a few interacing artifacts in the conversion. In other words, my script does not seem to be handling the source well.
Here is some info from the MediaInfo log:
Inform : MPEG-TS: 9.20 GiB, 2h 13mn
InternetMediaType : video/H264
Codec : AVC
Codec profile : Main@L4.0
Codec settings : CABAC / 4 Ref Frames
Codec_Settings_RefFrames : 4
Width : 1920
Height : 1080
Colorimetry : 4:2:0
Color space : YUV
Scan type : Interlaced
Scan order : TFF
Scan order : Top Field First
Interlacement : TFF
Bits/(Pixel*Frame) 0.173
My avisynth script is:
tfm()
Vinverse() ***great for helping with combing artifacts***
DGAVCIndex is giving me an error, with an option to ignore it. I have never seen this before:
Found NALU Type 13, len 2 undefined, ignore NALU moving on.
Simple script?...yes. But this has generally worked for me. I have tried various deinterlacers...yadif, tdeint, TomsMoComp. These generally leave artifacts that tfm does not, and are not doing any better deinterlacing than tfm. On a side note, I use MPC with ffdshow to view these files, and ffdhsow is not deinterlacing them correctly. Cyberlink seems to do the best. Anyways...
Here is a snapshot of one of many areas where the deinterlacing is going array:
So, attached is a small clip of the source file for anyone interested. Any help is always appreciated.
http://www.sendspace.com/file/i092ci (just look for the sample.ts link)
Thanks in advance,
David
+ Reply to Thread
Results 1 to 21 of 21
-
-
Last edited by Cauptain; 26th Aug 2010 at 13:18.
-
why 720p60 instead of 720p50 ?
I haven't looked at your sample, but assuming it's a 1080i50 PAL source, you will get either blends (=blurry like your screenshot) or duplicate frames (= choppy) because you don't have enough temporal information if you try to convert it to 720p60 instead of 720p50. You're trying to generate more frames per second out of nothing . Your starting PAL field rate is 50 fields per second , not 60. So "PAL to NTSC" is not the easy part at all. It shouldn't be done.
Don't use dgavcindex for interlaced streams, it has well documented problems with PAFF streams (blocky decoding errors). Use DirectShowSource + ffdshow
TFM is a field matcher, not a deinterlacer
If you want a great deinterlacer, try tempgaussmc_beta2 , way slower, but much fewer artifacts than yadif or the others listed above , yadifmod+nnedi2/3 is a decent alternative as wellLast edited by poisondeathray; 26th Aug 2010 at 13:13.
-
[QUOTE=poisondeathray;2013170]
Don't use dgavcindex for interlaced streams, it has well documented problems with PAFF streams (blocky decoding errors). Use DirectShowSource + ffdshow
/QUOTE]
How could I use? Could you give me an example? I found it very interesting.
Claudio
-
DirectShowSource("video.ts")
You need haali media splitter and ffdshow enabled to decode avc/h.264 -
Well I feel stupid . I assumed for some reason that he meant 720p60 video, not a film rate 25p=>24p conversion
The deinterlacing and DGAVCIndex comments still apply -
Weel I should have looked at your sample in the first place
It has field blends
I would use DSS2() because it's frame accurate . I think you could use a bobber, and srestore(23.976) to return the film rate and get rid of some of the blends
EDIT: it appears there are stream errors too , that might be another reason why dgavcindex "barfed" -
Yes, the field blends are why you're getting the "double images" or ghosting
DSS2("sample.ts")
YadifMod(order=1,mode=1,edeint=nnedi2(field=3))
SRestore(23.976)
AssumeFPS(24000,1001)
You can still process the audio in eac3to , but this script will return the film rate and get rid of most of the blends (no need to go through mkvmerge or timecodes) . If you want to resize, you can add LanczosResize(1280,720) after srestore -
This is why I needed expert help. Could you kindly help me with how to use haali and ffdshow in lieu of dgavcindex. In addition, if I am to use dss2, could you give me an example of what the script might look like. I have done things the same way for too long.
Thanks again,
David -
for haali , you just install it , and ffdshow you enable h.264/avc in the ffdshow configuration (it should be enabled by default anyway, but double check)
DSS2() is a replacement for directshowsource, and comes with haali splitter package. Put avss.dll copy in your avisynth/plugins folder and it will autoload
You can find yadifmod, nnedi2 and srestore easily from searching. If you can't just ask , and I'll point you in the right direction
I think I posted before your last post, because I posted example script there
Note, the deblending filters aren't perfect - they catch most of the blends, but some aren't "fixable" . -
-
do you mean the real video or the sample.ts ?
can you preview the script to the end (e.g in vdub) ? or does it crash ? or try another GUI or encoder ?
you have errors in even that small sample, so I would guess you have them in the full video as well. I suspect it's a problem with your source and decoding it (even DGAVCIndex reported errors)
another options if you have nvidia card with cuda, is to use DGNVTools , but it's not free . It's basically replacement for DGAVCIndex, but updated and faster
You can try directshowsource() instead, but it's not very good (not frame accurate, so the deblending function won't work as well)
captauin's script just bob deinterlaces it - so it doesn't address the blended fields (also it's not a video souce, it's a film source) -
-
-
Then you shouldn't be posting about methods to resolve the problem because your method sucks. The framerate isn't 50fps but 23.976fps, and you don't have a fluidity that you will not believe. And you keep all the blending. And you make unnecessary colorspace conversions. All further degrade the video quality.
Test and then tell me what you think.
What is it with you and all the bold-faced pronouncements? -
Manono are you okay?
You fight with your boyfriend or something? Why you are stressed? He´s left you?
I posted a method that works for me in my recordings.
Normally use MPEG2 or AVI, I found that with AVC source is different. I posted about the comment/tip from Poison.
If you can see so much so, nor should answer anything.
I don´t help you, you did not ask. So you have no reason to be like a madwoman (no offense to woman).
A next time, try to help instead of getting me around here. I like women, you... I have my doubts.
Ps: My apologies to members who did not need see it.
ClaudioLast edited by Cauptain; 27th Aug 2010 at 18:47.
-
back to our regular scheduled programming...
Here is an oddity. I decided to just watch the .ts file today in all of it's low bitrate glory. Turns out, this is a strange one...Unlike every PAL film I have downloaded, this one is actually time-correct. In other words, it does not suffer from PAL-speedup. It is the exact length of the NTSC version at 23.976 fps. I won't even pretend to figure out why this is, but I am sure however this movie was coded for PAL broadcasting, whatever method they used probably explains why this .ts file is so glitchy. -
Similar Threads
-
Blu-Ray 1080i source > MKV, confused about deinterlacing
By Phat J in forum Video ConversionReplies: 0Last Post: 24th Aug 2011, 23:42 -
Somewhat experienced user who's hit a wall converting and editing MTS files
By The Realms of Gold in forum Video ConversionReplies: 3Last Post: 13th Nov 2010, 23:44 -
Question for experienced ISOBuster user
By jman98 in forum Newbie / General discussionsReplies: 14Last Post: 15th Apr 2010, 19:44 -
Deinterlacing 1080i TS (what am I doing wrong?)
By dafoomie in forum Video ConversionReplies: 2Last Post: 10th Nov 2009, 15:53 -
Blackmagic Intensity Pro, 1080i video, and deinterlacing
By lbgaus in forum Capturing and VCRReplies: 0Last Post: 15th Sep 2009, 00:44