Go Back    Forum > Digital Video > Video Project Help > Restore, Filter, Improve Quality

Reply
 
LinkBack Thread Tools
  #41  
07-24-2014, 11:16 AM
Zerowalker Zerowalker is offline
Free Member
 
Join Date: Dec 2011
Posts: 121
Thanked 5 Times in 4 Posts
Wait, are you able to Capture at 10bit?
Reply With Quote
Someday, 12:01 PM
admin's Avatar
Ads / Sponsors
 
Join Date: ∞
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
  #42  
07-25-2014, 07:55 AM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
Yes !
I have captured the SDI output from the TBC using a Blackmagic SDI to PCIE mini recorder card.
The footage is captured in Apple V210 lossless 10bit codec files which Blackmagic provide. I think they are about 120G/hr
There aren't many available alternative lossless 10bit codecs.
I am so far able to do some analog restoration using AVIsynth in 10bit resolution. But at this point in time I have not been able to do deinterlacing with QTGMC in 10bit.

I should point out that I am a bit of a newby to AVIsynth.

The best part is probably that the video converted directly to digital (12bit internally before proc amp) and not 8bit A to D then D to A and then A to D again if using an analog output TBC. It is also lossless and avoids using DV capture.
I also own a Leitch DPS575 with analog to SDI which also has a good TBC with a 3D comb filter for composite input.
This would be great for capturing Laserdiscs in native composite video, but I don't have any!
Reply With Quote
The following users thank WaxCyl for this useful post: msgohan (08-10-2014)
  #43  
07-26-2014, 12:43 AM
Zerowalker Zerowalker is offline
Free Member
 
Join Date: Dec 2011
Posts: 121
Thanked 5 Times in 4 Posts
Impressive, 10bit is as rare as it comes for capture cards.
Didn't think you could receive and capture 10bit in the consumer market at all.

I really would like to see the advantage compared to 8bit. But it's impossible as i can't display 10bit, and i think even to get that to work is Special software, certain graphic cards etc, really hope it becomes standard though, 10bit is like 16bit audio to 24bit, it's kinda overkill, but you get to the point where any rounding errors that appear are impossible to notice.

Well it sure is important to keep track of the setup.
I haven't got any TBC and just a capture card, but that's it.

No idea of SDI works, just heard about it, but could you link your TBC and Card, a bit interested in this "SDI/10bit"
approach

Though, isn't the video 8 bit originally, where does it process to 10bit, is it the TBC for you?
Reply With Quote
  #44  
07-26-2014, 02:40 AM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
I am capturing analog video so it is neither 8bit or 10bit. In theory analog video is "smoother" than digital video. In practice I would expect most modern analog sources (ie VCR or 8mm camcorder) to have internal digital processing (albeit at 10bit) internal to the chippery. Nevertheless the most lossless output available from an analog source is analog s-video or component!

8bit digital video is excellent for watching video, but when processing is required 10bit comes to the fore.This is especially so for source material that is too dark or over exposed and lacking contrast.

When significant tonal corrections are made 8bit video will show banding (only 256 shades or graduations)
When 10bit video is corrected there are 1024 shades available. This why most audio/video devices now use higher bit resolution internally in the processing chips.

10bit processing is significantly better for RGB><YUV conversion
MPC-HC can be modified for viewing 10bit video. Will post more later
Reply With Quote
  #45  
07-26-2014, 03:06 AM
Zerowalker Zerowalker is offline
Free Member
 
Join Date: Dec 2011
Posts: 121
Thanked 5 Times in 4 Posts
So analogue isn't in any "bit"?
Well guess it's the same as Audio there, there is no depth, we simply have it there as we can't produce infinite data.

Had no idea a VCR would digitize, are you sure about this?

I have some experience in Avisynth and processing, and yes as you say, when processing 8bit the rounding errors will stack and you will get banding (Similar to getting noise when working in 16bit with audio).

That however is simply fixed by upconverting an 8bit video to 16bit, with that you get a huge working ground and rounding errors are no longer an issue.

Then after you are done you can simply downconvert it to 8bit or 10bit.

I usually do this, and output 10bit to x264 so it can process the data directly (no, 16->8->10) conversions.
Reply With Quote
  #46  
07-26-2014, 04:42 AM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
To correct myself the 8bit video we are discussing is actually REC601 which means the luma is only 16-235 which is only 219 shades of grey!
http://www.blackmagicdesign.com/au/p...specs/W-DLK-06
for capture of SDI to PC
http://www.ensembledesigns.com/products/brighteye/be75
recommend this for TBC + A to D + audio capture. no other models suitable
It is very expensive brand new ($nearly $2K in Australia).
Got mine on Ebay for ~$250

This product has 10bit analog capture
http://www.blackmagicdesign.com/au/p...sity/techspecs
it will not work with tape sourced video as it does not have a TBC

Well I am certain that any decent video deck with DNR and or LTBC has AD and DA conversion.
OSD and antishake filters as well would be digital domain

My partial 10bit processing script

Code:
SetMTMode(5, 8)
FFvideosource("M:\_BE Capture\B1.avi", enable10bithack=true) # source is 10bit 720x576 interlaced 4:2:2
AssumeTFF(last)
dfttest() #NR
Autoadjust (auto_gain=false, high_quality=true, high_bitdepth=true, chroma_process=180) #Auto levels
f3kdb_dither( input_depth=16, keep_tv_range=true) 
#no more 10 bit after here
ConverttoYUY2(matrix="rec601", interlaced=true)
FixChromaBleeding(last) #
ChromaShift(last, c=0, V=2, U=0, L=0 ) 
QTGMC(Preset="SLOW", Sharpness=1.5, EdiThreads=4, ChromaEdi="NNEDI3", NNeurons=4, NoiseProcess=2, NoiseRestore=0.0, Denoiser="fft3dfilter", DenoiseMC=true, NoiseTR=2, Sigma=4.0)
nnedi3_rpow2(2, cshift="LanczosResize", fwidth=960, fheight=720)
ReSampleHQ(srcmatrix="TV.601", dstcolorspace="RGB32", dither=true, chroma_kernel="Spline36")# , RGB ready for edit
Reply With Quote
The following users thank WaxCyl for this useful post: msgohan (08-10-2014)
  #47  
07-26-2014, 08:06 AM
Zerowalker Zerowalker is offline
Free Member
 
Join Date: Dec 2011
Posts: 121
Thanked 5 Times in 4 Posts
Well then 10bit processing isn't a problem right?
I mean, as long as you can input it, you just need to upsample to 16bit (as you do), then do the processing, then downsample to 10 or 8bit in the final production.

Pretty much what you do with 8bit.
Reply With Quote
  #48  
07-26-2014, 01:56 PM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
It would be fantastic if Avisynth had native 16bit processing and all filters as well..
My analog tape collection is starting to get sticky and unplayable so the most important positive to all this is that the master captures are at least 10bit resolution, and ready for the future.
I would prefer if I could capture at 4:4:4 and 768 x 576 (4:3) and NOT D1 (720 x 576 anamorphic 5:4)
At least it is an improvement compared to lossy DV capture (4:2:1 PAL or 4:1:1 NTSC )
Reply With Quote
  #49  
07-26-2014, 02:46 PM
Zerowalker Zerowalker is offline
Free Member
 
Join Date: Dec 2011
Posts: 121
Thanked 5 Times in 4 Posts
There is no gain in 4:4:4. It's like taking a YV16 file and upconverting it to YV24, it makes no sense, it will just look identical for larger size.

768x576 is no use either.

The information from 720->768 doesn't exist, that's why the Aspect Ratio exist to "guess" it.
You can either capture it to 4:3, resize it to 4:3 or let the playback to it for you.
All will be the same (Except for the resizer of use).
Reply With Quote
  #50  
07-26-2014, 03:20 PM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
You are forgetting that this is analog video. None of these parameters apply natively. This is not "upscaling".
4:2:1 captures does seem to exacerbate the already poor analog chroma signal. 4:2:2 is excellent admitedly.
I am simply trying to capture analog video without adding further losses in the digital domain.
Analog video does not have a horizontal pixel resolution but I think it is equivalent to about 420 (lines) for Hi8.
The D1 digital video standard is about 20 years old!
Reply With Quote
  #51  
07-26-2014, 03:34 PM
Zerowalker Zerowalker is offline
Free Member
 
Join Date: Dec 2011
Posts: 121
Thanked 5 Times in 4 Posts
True, but for me using PAL i am not sure how it will look.
As every other line in PAL is "useless", meaning it's pretty much identical to YV12.
So the solution is to discard those lines of any chroma information.

So capturing in 4:2:2 would be a waste for it.

NTSC however, is closer to YV16, so the information there is another story.

As for the "pixel resolution".

If i remember correctly nothing should come even close to 720x576 except DVD.
And i don't think anything actually captures 768x576. I know one of my cards can, but from my tests, it looks pretty much identical, so it's probably just stretching the data (like it does with it anyway as there is no accurate resolution, you just have to capture larger than it really is to not miss information)
Reply With Quote
  #52  
07-26-2014, 08:52 PM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
I am in a PAL country as well.
My work flow until now for creating viewable video has so far been to keep the video 50p.
Bluray supports 50p at 1280 x 720
After QTGMC and editing (all at 50p) I use nnedi3 to resize to 1280 x 720 and encode to H264 at
25Mb/s ( probably near lossless considering the source material). Don't discard fields after deinterlacing!
The net result of this workflow is that absolutely no data is wasted including the 50fps temporal infomation of interlaced PAL. Slow motion works really well from this.
Hi bitrate processing would eliminate the only "loss" - converting YUY2 to RGB.
The results are truly amazing especially when I tell interested friends Yes! it was analog Hi8
Reply With Quote
  #53  
07-26-2014, 11:55 PM
lordsmurf's Avatar
lordsmurf lordsmurf is offline
Site Staff | Video
 
Join Date: Dec 2002
Posts: 14,052
Thanked 2,554 Times in 2,172 Posts
Quote:
Originally Posted by Zerowalker
There is no gain in 4:4:4. It's like taking a YV16 file and upconverting it to YV24, it makes no sense
768x576 is no use either.
You beat me to it. This is correct.

- Did my advice help you? Then become a Premium Member and support this site.
- For sale in the marketplace: TBCs, workflows, capture cards, VCRs
Reply With Quote
  #54  
07-27-2014, 02:39 AM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
You are both correct : upscaling to 4:4:4 is a waste of time.
I was NOT proposing this.
Any analog capture, though, that is less than 4:4:4 is lossy.
You can't throw out 50% of your chroma data and call it lossless!

Sampling the video horizontally at 768 pixels (compared to 720 and then interpolating to 768) is a small improvement. It means aliasing artifacts are moved further out of the passband.

What I was hoping was that someone more advanced than I at Avisynth would criticize my avs script above.
The instructions for using Avisynth seem to be spread across the entire internet!

If, suppose, my job was to capture for posterity the last remaining Wax Cylinder(!) recording of a famous singer for the National Library. If it was up to me and I had a choice to capture at 16bit/44.1 or 24bit/96Khz I would go the latter- And I would manufacture a high quality magnetic transducer for the pickup on the player rather than sticking microphone into the horn!
The 10bit video captures I have so far made are quite visibly superior to my previous DV captures.
You really must try it - it is truly worthwhile.
Reply With Quote
  #55  
07-27-2014, 04:41 AM
lordsmurf's Avatar
lordsmurf lordsmurf is offline
Site Staff | Video
 
Join Date: Dec 2002
Posts: 14,052
Thanked 2,554 Times in 2,172 Posts
Why use
Code:
FixChromaBleeding(last)
instead of
Code:
ChromaShift(C=x, L=x) # align chroma over luma
where x= the positive/negative pixels to shift.

- Did my advice help you? Then become a Premium Member and support this site.
- For sale in the marketplace: TBCs, workflows, capture cards, VCRs
Reply With Quote
  #56  
07-27-2014, 07:14 AM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
Thanks for looking!
I did use chromashift in the line above it.
FixChromaBleeding has a different function to Chromashift.
I have also tried FixChromaBleedingMod but it does not always work and it sometimes causes weird blocking
I have since found a better way to use FixChromaBleeding:

Code:
ChromaShift(last, c=0, V=0, U=-0, L=-4) #
FixChromaBleeding()
FlipHorizontal()
FixChromaBleeding()
FlipHorizontal()
I have some 20YO video (not mine)that was copied from a Lo8 cam to my SVHS.
The original tape was stolen with the camcorder. This is from a DV capture from the SVHS and I have been attempting to repair the Chroma flare. Check out my B4 and after. I only used neat video for the Chroma.
Sorry I managed to grab a different frame and yes, the chroma gain is too high, but it it does emphasise the improvement.


Attached Images
File Type: jpg DVclipwJustBobDub.jpg (72.4 KB, 38 downloads)
File Type: jpg FullSynthQT+Neat.jpg (70.2 KB, 33 downloads)

Last edited by WaxCyl; 07-27-2014 at 07:26 AM. Reason: Added last line
Reply With Quote
  #57  
07-27-2014, 07:26 AM
lordsmurf's Avatar
lordsmurf lordsmurf is offline
Site Staff | Video
 
Join Date: Dec 2002
Posts: 14,052
Thanked 2,554 Times in 2,172 Posts
We need to add some "do not buy" advice to this thread.
For example, the Leitch DPS-475/575 -- http://www.digitalfaq.com/forum/vide...ade-tbc-2.html

- Did my advice help you? Then become a Premium Member and support this site.
- For sale in the marketplace: TBCs, workflows, capture cards, VCRs
Reply With Quote
  #58  
07-27-2014, 07:42 AM
WaxCyl WaxCyl is offline
Free Member
 
Join Date: Jan 2014
Posts: 55
Thanked 8 Times in 8 Posts
The DPS475 is NTSC only and could be a problem for some. I have a DPS575. When the TBC is engaged it has only 8bit resolution but SDI 10bit or analog output
I am unable to see any difference in the output quality compared to the BE75.
As I have said previously the benefit of 10bit will mostly show when making adjustments that will otherwise cause banding.
The unit is monstrously big and I really can't recommend it: unless someone has a large collection of laserdiscs that they want to capture with superior quality - it has a 3D comb filter for composite input
Reply With Quote
  #59  
07-27-2014, 09:51 PM
lordsmurf's Avatar
lordsmurf lordsmurf is offline
Site Staff | Video
 
Join Date: Dec 2002
Posts: 14,052
Thanked 2,554 Times in 2,172 Posts
Quote:
Originally Posted by WaxCyl View Post
As I have said previously the benefit of 10bit will mostly show when making adjustments that will otherwise cause banding.
For a proc amp, that would make a small difference, yes.

- Did my advice help you? Then become a Premium Member and support this site.
- For sale in the marketplace: TBCs, workflows, capture cards, VCRs
Reply With Quote
  #60  
08-08-2014, 03:03 PM
msgohan msgohan is offline
Free Member
 
Join Date: Feb 2011
Location: Vancouver, Canada
Posts: 1,323
Thanked 336 Times in 277 Posts
Quote:
Originally Posted by WaxCyl View Post
Any analog capture, though, that is less than 4:4:4 is lossy.
You can't throw out 50% of your chroma data and call it lossless!
You can when none of your sources come close to having full chroma resolution. Do you really think your home recordings using consumer equipment have more chroma resolution than, say, a 4:2:2 studio master made from a 35mm film scan in 1990 or 2005 simply because your recordings are stored on analog tape?

The process of color-under recording itself is a form of lossy analog compression that massively cuts down the chroma resolution and is the reason for color bleed on tapes in the first place.

Quote:
Sampling the video horizontally at 768 pixels (compared to 720 and then interpolating to 768) is a small improvement. It means aliasing artifacts are moved further out of the passband.
Basically any decent capture chip since like 1998 internally samples at 1440 or greater and downsamples to the specified output resolution.

It's worth noting that unless you turn off the TBC in your S-VHS or Hi8 player, you've already converted your video to 8-bit 4:2:2 (or less) before it has the chance to reach your 10-bit converter. So you are giving yourself more headroom for your second A-D step, but the signal has already been limited. Even the TBCs in Betacam SP decks were only 8-bit.

Last edited by msgohan; 08-08-2014 at 03:30 PM.
Reply With Quote
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
Done with wasting my time trying to capture VHS with a low budget VCR! jrnyhead Capture, Record, Transfer 13 02-14-2012 04:44 PM
How to best perform overscan correction? dyfan Restore, Filter, Improve Quality 8 02-27-2010 06:23 PM
Eurorack UB802 for audio correction, help using it lordsmurf Restore, Filter, Improve Quality 0 01-15-2010 04:04 PM
Recording time on DVD-R? genebender Blank Media 1 02-08-2009 12:14 PM

Thread Tools



 
All times are GMT -5. The time now is 05:13 PM