SD video becomes dull when upscaling to HD

[Bencuri]

[QUOTE=hello_hello;2740904]

I know what you mean about stretched looking faces in the screenshot you posted but I think it's just a perspective thing due to the angle. It's hard to tell but at least at 16:9 it doesn't look like they all have eating disorders.

I experimented with the various codes you and other submitted, and I noticed that for some reason your code results in a clearer image in the end than the one jagabo added, even if the image is more streched, thus the screen size is bigger. Maybe because of Resize8. That is odd because I would have expected that the more you stretch this the more blurred it will be.

However I decided I would go with a ratio closer to 4:3, because of those parts that I mentioned, where the ratio might have been different and would become too overstretched. But I noticed I cannot stretch the image vertically if I try to enter the target values in Cropresize. The result is zooming in instead of stretching. What I can do is to add Crop(0,70,-4,-70) to the beginning of the code to stretch the image vertically before any resizing and deinterlacing. I am not sure it is the proper way though. Should I acheive it with Cropresize instead, to gain more sharpness? I am not sure, just asking, have no idea how the two crop method differ.

[Bencuri]

In the early 2000's the UK and perhaps some other places experimented with the 14:9 (1.555) aspect ratio
as a way to "ease" the transition of 4:3 > 16:9. Perhaps this is what happened here.

Not sure. Maybe. But it doesn't matter now because the size varies definately along the clip, so better to find a compromise now by trial and error not by looking at the standards.

[jagabo]

the size varies definately along the clip, so better to find a compromise now

Or resize/pad sections to get the right aspect ratio for each.

[Bencuri]

the size varies definately along the clip, so better to find a compromise now

Or resize/pad sections to get the right aspect ratio for each.

That is the delicate solution but out of time now, it would take too long to correct the whole DVD like that.

[Bencuri]

Finally I chose 14:9 as a compromise for the latest clip I presented, most other clips on the DVD are 4:3 so it is a good compromise. I went with Resize8 because it seemed to me it does a better job than SplineResize. I skipped Dehalo from the script, because it seems to blur the already blurry image even more, so in the end it is just worse than without it. But there seems to be a little problem. Not the end of the world, but I wonder if it could be improved?:

If you check the segment I attached, especially the part where the guy in the black dress and hat moves his head, that movement has a quite noticeable judder. I checked it, and you can spot that judder in the original mpg as well, but overally it is less noticeable throughout the whole clip. With dehalo aplha, the upscaled clip has a slight judder as well, but without dehalo alpha, it becomes stronger. I guess it is because that is a usual judder, that is emphasized as the image is sharpened. Is there any way to lessen that judder? If not, no problem. It is still better with juddering than with a blurry image.

[jagabo]

You're talking about the shot from 3 to 5 seconds? I don't see any unusual judder. Just the normal 50 fps video to 60 fps monitor judder (10 little jerks per second, barely noticeable even on smooth panning shots). If you step through frame by frame there is equal motion between frames; no duplicate frames.

[Bencuri]

You're talking about the shot from 3 to 5 seconds? I don't see any unusual judder. Just the normal 50 fps video to 60 fps monitor judder (10 little jerks per second, barely noticeable even on smooth panning shots). If you step through frame by frame there is equal motion between frames; no duplicate frames.

That may well be the issue as my screen is also 60p.

Here is a better example anyway, maybe it will be more obvious now: the guy in white between 0:05 and 0:07. You can see it very emphasized during his movements.

If I add the dehalo function and let it blur the image, it is a bit better, not that emphasized, but then the problem is that the image becomes too blurry.

Not a very serious issue though. I just noticed it sometimes in the past as well and wondered if you can do anything with it. If not, that is still not the end of the world.

[jagabo]

I don't see anything wrong in that section of the video. There is a big bitrate peak at the 5 second mark (an I frame). Maybe you're having decoding problems?

[Bencuri]

I don't see anything wrong in that section of the video. There is a big bitrate peak at the 5 second mark (an I frame). Maybe you're having decoding problems?

What I mean is that kind of flicker that you see in case of low framerate videos. But it is only on the figure not the whole image. But it can be becuase of the 60p monitor indeed. But no problem then. It is not that serious. If you see nothing, it doesn't worth to worry about it.

[jagabo]

Adding some motion blur with QTGMC() might make it look a little smoother. Or using RIFE() to convert the video to 60p -- but that will add distortions to some parts of the video.

[Bencuri]

Adding some motion blur with QTGMC() might make it look a little smoother. Or using RIFE() to convert the video to 60p -- but that will add distortions to some parts of the video.

Can you suggest something? I see the parameters there and have some clue about shutter speeds, but I see you need to specify the blur in the source (ShutterAngleSrc), I have no idea how to determine that. I don't have too much experience.

[jagabo]

Motion blur with QTGMC:

LWLibavVideoSource("Obligatoire2_0_260_268.mp4", cache=true, prefer_hw=2) 
QTGMC(Preset="slow", InputType=1, ShutterBlur=1, ShutterAngleSrc=30, ShutterAngleOut=330, FPSDivisor=1 )

Motion interpolation to 60p with RIFE:

function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=9, fps_num=num, fps_den=den, sc=true, sc_threshold=0.10)
    z_ConvertFormat(pixel_type="YUV420P8", colorspace_op="rgb:709:709:f=>709:709:709:l")
}

LWLibavVideoSource("Obligatoire2_0_260_268.mp4", cache=true, prefer_hw=2) 
SmoothFPS_RIFE(60000, 1000)

[Bencuri]

I have insterted the blurring script to AVSPmod, but strangely it is causing combing. Why is that? If I remove the blurring part, only leave QTGMC(), the combs disappear.



[Attachment 80330 - Click to enlarge]


EDIT: Okay I have found if I remove everything from that line only leave the 3 shutter parameters, the combing stops.

I will need to experiment with it, but one thing I do not understand, why 330 shutter and not 180. What is the logic in selecting so high value right away?

[jagabo]

Okay I have found if I remove everything from that line only leave the 3 shutter parameters, the combing stops.

I was starting with your progressive sample so it didn't need deinterlacing. You are starting with an interlaced video so it needs deinterlacing.

I will need to experiment with it, but one thing I do not understand, why 330 shutter and not 180. What is the logic in selecting so high value right away?

I exaggerated the effect to make it very visible. Use whatever amount suits you.

By the way, with motion blur you still have the fundamental problem of 50 fps needing to be displayed on a 60 hz monitor. Ie,4 frames will be displayed for 1/60 second then the 5th will be displayed for 2/60 of a second -- 10 little jerks every second. The motion blur just makes it a little less visible.

[Bencuri]

Okay I have found if I remove everything from that line only leave the 3 shutter parameters, the combing stops.

By the way, with motion blur you still have the fundamental problem of 50 fps needing to be displayed on a 60 hz monitor. Ie,4 frames will be displayed for 1/60 second then the 5th will be displayed for 2/60 of a second -- 10 little jerks every second. The motion blur just makes it a little less visible.

No problem. I decided not to change this because at my location there is 50hz standard, so better leave it on 50p.

The motion blur does improve it. Easier for the eye to watch. But need to tweak it seriously, as the story is the same again: bad quality blurry parts just get blurrier. But I am trying to find a good compromise.

What I do not understand is: do you need to keep the framerate in mind in case of shutter angle? I mean, the standard is 180°, I know it, but what if you want to play with less blur/more blur? I thought you can just decrease and increase it freely and randomly: 180 - 200 - 270 etc. But what I am not sure about: do you still need to consider specific fractions when changing the angle? Similarly to the rule that shutter speed should be a multiple of the framerate or close to it. Or in case of angle you just need to check the blurriness, no math matters?

[jagabo]

Shutter angle in QTGMC is inherently relative to the frame rate. 180 degrees is halfway between frames whatever the frame rate is.

[Bencuri]

Shutter angle in QTGMC is inherently relative to the frame rate. 180 degrees is halfway between frames whatever the frame rate is.

I see but when I try different shutter angle, do I need to keep in mind any ratio relative to 180? Just like in case of shutter speed: in case the framerate is 25, as I know the recommended is to select multiple amounts of that (1/25, 1/50, 1/100.). Do I need to follow such a rule for shutter angle? For example need to select: 180 or 90 or 270? Or It doesn't matter even if I select something like 171 or 233, I just need to decide if I want to sharpen or blur, and increase or decrease as much as I want?

[jagabo]

You can select any shutter angles you like. The larger the shutter angle the more motion blur you get.