VideoHelp Forum

Our website is made possible by displaying online advertisements to our visitors. Consider supporting us by disable your adblocker or Try ConvertXtoDVD and convert all your movies to DVD. Free trial ! :)
+ Reply to Thread
Page 4 of 4
FirstFirst ... 2 3 4
Results 91 to 105 of 105
Thread
  1. Originally Posted by johnmeyer View Post
    I don't see any detail reduction in the Sony DV, other than what is caused by the macroblocking. Here is a link directly to the images, so you can download them directly:
    It's very clear in the video samples

    The trend is macroblocking but more detail retention with Mainconcept, less macroblocking but more blurring and detail loss with Sony. It's easy to see that the details like pant folds are gone, textures like granite on the walls, ashphalt road textures are reduced . It looks like Sony is applying a NR pre compression .

    Of course, to your average viewer, you won't notice these sorts of things in motion
    Quote Quote  
  2. Originally Posted by poisondeathray View Post
    Also, the ut sample was BFF
    Wow, you are right. My editing program (Vegas) reported it (the Lagarith version) as TFF. MediaInfo punts and doesn't report anything. I just checked using a separatefields() script, and it is definitely BFF.

    I just re-did the render and field/bob capture, and it looks identical when I do it using BFF: same exact artifacts.

    I also just did a render using the free Cineform/GoPro codec set to "medium" quality. It produces almost zero artifacts, and is only 10% larger than the DV codec. Something to think about, depending on your workflow.

    Of course the original intent of this thread was to capture VHS analog video, and much of the capture hardware is hard-wired to work with DV, and it works so well in so many other ways that I still recommend it for capturing old analog video.
    Last edited by johnmeyer; 9th Jun 2015 at 11:35. Reason: added "Lagarith version"
    Quote Quote  
  3. The lagarith is 4:2:2 also - so no problem there, false alarm. Interesting that it was larger than the ut video version, that almost never happens unless there are duplicates and the null frame option is selected

    Vegas treats lossless codecs like lagarith as RGB. The DV encoded versions look like the range was expanded

    I agree cineform is a great option, for other scenarios too
    Quote Quote  
  4. Originally Posted by poisondeathray View Post
    The trend is macroblocking but more detail retention with Mainconcept, less macroblocking but more blurring and detail loss with Sony. It's easy to see that the details like pant folds are gone, textures like granite on the walls, ashphalt road textures are reduced . It looks like Sony is applying a NR pre compression.
    It is always interesting to look at these things. I see what you are seeing: the flat walls of the columns show less detail with the Sony encoder. However, the vertical "fluting" in those columns, on their left edge, holds up much better in the Sony version than the MainConcept.

    The more I look at each encode, comparing them to the "original" lossless, the more I see parts of the image where one encoder does a better job, and the other does a worse job. For instance, since we're looking at the left side of the image, go to the right of the woman where a man in sunglasses is smoking a cigarette, and then look at the bald head on the man behind him. The details on that head and the hairline get degraded substantially by the MainConcept codec, but far less by the Sony DV codec.

    As always, the question is how it looks when you are sitting in your couch watching the result. If I look at first-generation 1080p AVCHD video (straight out of the camera), frame-by-frame (which I just had to do when trying to capture a still image for a runner), it is amazing how horrible each frame looks when analyzed as a still, even though the video looks remarkably sharp when viewed in motion.
    Quote Quote  
  5. Originally Posted by poisondeathray View Post

    I agree cineform is a great option, for other scenarios too
    As is Grass Valley (Canopus) HQX IMHO. Used with their free AVCHD2HQ conversion utility it produces excellent intraframe intermediate files to work with, both for SD and HD (and even 4K!) and with pretty reasonable files sizes (especially when the 'online standard' resolution is selected).

    Desigend to work originally with Edius of course, but now freeware and usable outside Edius as well...
    Quote Quote  
  6. Banned
    Join Date
    Oct 2014
    Location
    Northern California
    Search PM
    Originally Posted by johnmeyer View Post
    As always, the question is how it looks when you are sitting in your couch watching the result. If I look at first-generation 1080p AVCHD video (straight out of the camera), frame-by-frame (which I just had to do when trying to capture a still image for a runner), it is amazing how horrible each frame looks when analyzed as a still,...
    That is not just the fault of compression.

    Most consumer level 1080p camcorders don't really capture 1920x1080 pixels.
    Quote Quote  
  7. Some frames are better than others, this is true for all compression tests in general. But the trend I mentioned is very strong here. It's just one sample, but from what I see I'm willing to bet money it will be very similar on other samples. I have quite a bit of experience evaluating codecs - normally I would refrain from making a broad statement from 1 sample, but when I see something this clear it's a very strong predictor. It's funny, because this is almost a 180 on what Mainconcept does with their AVC codec. That has weak texture retention on flat areas, dark areas

    The question I had was how "bad" are the artifacts and losses. It definitely will be satisfactory for some people in some situations, but definitely not for others. I just wanted some quantification or hard examples
    Quote Quote  
  8. Member
    Join Date
    Oct 2013
    Location
    Germany
    Search Comp PM
    Originally Posted by 2Bdecided View Post
    Anyone who can see any difference between DV and lossless is going to be utterly utterly horrified at DVD-compliant encoding of challenging interlaced video content, viewed with the same level of criticism.
    Speaking of it this is the best I could get out of the sample while maxing out the DVD specs at 9600 CBR (so there's still room for stereo audio). Not surprising with this particularly extreme kind of material Procoder produced the least artefacts.

    Like vaporeon800 I used SeparateFields() and LanczosResize(width*2,height*2) for these screenshots.

    Click image for larger version

Name:	Lossless_2.png
Views:	145
Size:	974.2 KB
ID:	32130
    Click image for larger version

Name:	MPEG2_2.png
Views:	149
Size:	923.8 KB
ID:	32135

    Click image for larger version

Name:	Lossless_1.png
Views:	127
Size:	902.1 KB
ID:	32131
    Click image for larger version

Name:	MPEG2_1.png
Views:	127
Size:	874.0 KB
ID:	32134

    Click image for larger version

Name:	Lossless_3.png
Views:	129
Size:	995.6 KB
ID:	32132
    Click image for larger version

Name:	MPEG2_3.png
Views:	133
Size:	978.9 KB
ID:	32133



    I also tried HCenc and CCE SP3. I confirm, no matter what I tried, HCenc pretty much fails at this sample.
    Here's a HCenc screenshot (same frame as the 2nd one above).

    Click image for larger version

Name:	MPEG2_HC.png
Views:	104
Size:	884.1 KB
ID:	32136
    Image Attached Files
    Last edited by Skiller; 9th Jun 2015 at 13:22.
    Quote Quote  
  9. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Thank you Skiller. It's what I've been seeing for years.

    HCenc isn't very good with high quality detailed interlaced camcorder footage either. TBH MPEG-2 itself isn't really; the "better" encoders start to throw away details and sharpness to avoid artefacts. If you're going to look closely at individual frames, you're going to be very disappointed.


    If you're going to print out screen captures to photographic paper, or upscale your VHS captures to Full HD, burn them to high-bitrate BluRay, and then press pause to look at individual frames, then maybe you'll see the advantages of lossless capture. But if your VHS captures are heading to DVD, I just can't see how you'll ever see the difference. DV is far from the weakest link in that chain.

    Cheers,
    David.
    Quote Quote  
  10. Skiller: Excellent. What about compounding losses from a trip through DV prior to MPEG-2? I've uploaded the Cedocida sample, if you still have your Procoder settings available.

    johnmeyer: Thanks for those. Not to be a bother, but any chance you can work around the RGB crushing/clipping issue noted by poisondeathray to make new samples (in BFF)? Perhaps if you write out the Lagarith sample to uncompressed YUY2 using VirtualDub, Vegas would treat the uncompressed one appropriately?

    Originally Posted by poisondeathray View Post
    Interesting that it was larger than the ut video version, that almost never happens unless there are duplicates and the null frame option is selected
    I've found that Ut Video is smaller any time I've tested on my own VHS captures. It turns out I used an outdated Lagarith version (1.3.20), but the newest (1.3.27) only reduces the size of this clip by 2.2MB. It's still almost 12MB larger than the Ut Video.

    Originally Posted by 2Bdecided View Post
    upscale your VHS captures to Full HD
    Aren't most of us watching on 1080p screens even if we burn to DVD specs? I know some people keep a CRT just for viewing old material, but I don't think it's typical.
    Last edited by vaporeon800; 14th Jun 2015 at 14:30.
    Quote Quote  
  11. Member
    Join Date
    Oct 2013
    Location
    Germany
    Search Comp PM
    Originally Posted by vaporeon800 View Post
    What about compounding losses from a trip through DV prior to MPEG-2?
    Good idea! Thanks for the DV sample.


    Since Procoder needs YUY2 input I requested Cedocida to decode the DV source to YUY2 via pixel_type="YUY2" in AviSource.


    Here's a roundup:


    Lossless
    MPEG2 encoded from Lossless
    DV (Cedocida)
    MPEG2 encoded from DV

    Lossless
    MPEG2 encoded from Lossless
    DV (Cedocida)
    MPEG2 encoded from DV

    Lossless
    MPEG2 encoded from Lossless
    DV (Cedocida)
    MPEG2 encoded from DV



    Now I wasn't really expecting much of any difference between the MPEG2 sample that was encoded from the lossless original and the one that went through the DV trip (except for the artefacts to "look different"). However, especially in the first set of screenshots the one that was encoded from the lossless original does exhibit quite a bit less mosquito noise – have a look at the building in the background at the top of the picture, the skyline and the building on the left.
    This made me suspicious about really using the same settings in Procoder that I used a few days ago (there aren't that many settings in Procoder anyway). So I double checked by encoding the lossless original again today using the settings I just used for the DV one... and I got bit idential output compared to what I got a few days ago, so any difference is really down to the DV encoding!

    In the second set of screenshots there is more mosquito noise around the guard's stick (sorry, don't know what's it called).

    In the third set there is blocking in the DV sample that made it perfectly all the way into the MPEG2 encoding of it and is barely there in the losslessly sourced MPEG2 encode! Look at the gate in the top right corner.

    So, there's a difference, even when encoding for DVD. I'm surprised. But let's put it into perspective: we are looking through the magnifying glass at a single field we would normally see for 0.02 seconds. Also, this is a very very hard to encode video, the differences may be amplified because of that, or maybe not. Something else to test.
    Image Attached Files
    Last edited by Skiller; 15th Jun 2015 at 10:10.
    Quote Quote  
  12. Video Restorer lordsmurf's Avatar
    Join Date
    Jun 2003
    Location
    dFAQ.us/lordsmurf
    Search Comp PM
    Originally Posted by vaporeon800 View Post
    Aren't most of us watching on 1080p screens even if we burn to DVD specs? I know some people keep a CRT just for viewing old material, but I don't think it's typical.
    I keep some CRTs on hand. One can be calibrated very close to proper values.

    The rest are just here because of "ain't broke, don't fix (replace) it". Great for the garage, the kitchen, spare room, etc.
    Quote Quote  
  13. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Originally Posted by vaporeon800 View Post
    Originally Posted by 2Bdecided View Post
    upscale your VHS captures to Full HD
    Aren't most of us watching on 1080p screens even if we burn to DVD specs?
    I was thinking the way to actually see a difference is to capture to lossless, upscale, and then encode to BluRay specs. I was thinking this could be (slightly) visibly better than capturing to DV, upscaling, and encoding to BluRay specs.

    I was thinking that the DVD encoding stage would hide the difference. However, I'm willing to admit I was wrong. The first comparison in Skiller's post shows some of the mosquito noise from the DV encoding making it through to the MPEG B-frame.

    In the second comparison, there's blocking on the guard's arm in the MPEG-2-from-DV encode that's absent in the MPEG-2-from-lossless encode. This is especially strange because that blocking isn't visible in the DV version, but obviously something else in the DV version is upsetting the MPEG-2 encoding - probably the increased noise overall.


    I'd rather watch the DV version than the MPEG-2-from-lossless version (the damage done by MPEG-2 encoding on B-frames is much larger than the damage done by DV encoding), but if you have to create a DVD, this example shows it can be visibly better to create it from lossless than DV.

    I've seen this effect with high quality sources, but this is the first time someone has proven it with a VHS source. Thank you Skiller. In thinking this couldn't matter, I was wrong!

    I have to admit though, even when I know what to look for, I can't see the difference when I'm watching the video normally.

    Cheers,
    David.
    Last edited by 2Bdecided; 16th Jun 2015 at 05:48.
    Quote Quote  
  14. Member
    Join Date
    May 2014
    Location
    Memphis TN, US
    Search PM
    Originally Posted by Skiller View Post
    In the second set of screenshots there is more mosquito noise around the guard's stick (sorry, don't know what's it called).
    It's called a baton. http://www.merriam-webster.com/dictionary/baton
    - My sister Ann's brother
    Quote Quote  
  15. Perfect, Skiller, thank you!

    Originally Posted by 2Bdecided View Post
    I have to admit though, even when I know what to look for, I can't see the difference when I'm watching the video normally.
    I'm certain I could pick out which is which in a "blinded" full-screen, full-speed playback test... if I'm allowed to stare at the blocking/mosquitoes around the burned-in time display.
    Quote Quote  



Similar Threads