change the very low resolution of video to high resolution

[idi0t]

hi
i have a very old video which is recorded on old low revolutionizing video camera. i just have the video in .webm format.i want to convert it to high resolution video.
please tell me any way to convert it.even if i have to convert it fame by frame and remake it.
thank you very much
keep in mind the video has very low resolution because it is made in 70s. i think so.

[LMotlow]

You can't convert low resolution to high resolution. All you can do is make the frame bigger. It will still look like crappy webm low resolution.

[idi0t]

can i remake it in adobe effects or any other software.
i mean just manually make the video just like animation.

[Cornucopia]

That could mean a LOT of things. We can't read your mind, so please be specific. And thorough.

As a rule (which cannot be broken, but sometimes bent), you cannot turn low rez into high rez (certainly not without putting in so much effort as to be equivalent to just creating new high rez material). Same thing for low fi to high fi, stills to motion, b&w to color, mono to stereo, 2d to 3d, stereo to surround...

AE and others can DO video & animation, but some forms (stop-motion, 3d cgi, cell, e.g.) are better served with different apps.

Scott

[johnmeyer]

I found the previous two answers unintentionally misleading. Both say that you cannot turn a low resolution video into a high resolution video. As written, both statements are wrong. You can easily turn a 320x40 video into 1920x1080 by using a "resize" command, either in AVISynth or in your NLE.

However ...

What both people above are trying to say is that the resulting video won't contain any new details nor will it look any sharper. The re-sizing operation will reduce the jaggies on diagonal lines, and depending on how you do the re-sizing, may also be able to reduce other artifacts that sometimes accompany low-resolution video. But it will still look just like the original video, and will not have any new detail, nor will it look any sharper.

[pandy]

I found the previous two answers unintentionally misleading. Both say that you cannot turn a low resolution video into a high resolution video. As written, both statements are wrong. You can easily turn a 320x40 video into 1920x1080 by using a "resize" command, either in AVISynth or in your NLE.

There is nothing misleading in those answers - low resolution video will stay low resolution video even upscaled to UHD or higher.

However ...

What both people above are trying to say is that the resulting video won't contain any new details nor will it look any sharper. The re-sizing operation will reduce the jaggies on diagonal lines, and depending on how you do the re-sizing, may also be able to reduce other artifacts that sometimes accompany low-resolution video. But it will still look just like the original video, and will not have any new detail, nor will it look any sharper.

However there is possibility to synthesize some higher resolution data with help of statistical signal processing so called super-resolution http://en.wikipedia.org/wiki/Super-resolution_imaging. And resizing usually not reduce perceived jaggies (special class of resizers must be used) - worse it may emphasize low resolution limitations and perceived image quality will be worse than perceived quality of the low resolution source.

[idi0t]

Thanks for all your answers.
i have just attach a image of an anime character.
i want to turn my video to look like anime world characters.
i hope you got it.
if it is possible tell me please.


[Attachment 42135 - Click to enlarge]

[pandy]

Thanks for all your answers.
i have just attach a image of an anime character.
i want to turn my video to look like anime world characters.
i hope you got it.
if it is possible tell me please.

Difficult task...

Waifu2x for example in avisynth https://github.com/sunnyone/Waifu2xAvisynth

But first you should learn on it:
https://forum.videohelp.com/threads/372157-New-upscaling-algorithm-waifu2x
http://forum.doom9.org/showthread.php?t=173244
http://forum.doom9.org/showthread.php?t=172390
http://forum.doom9.org/showthread.php?t=172151

+

http://forum.doom9.org/showthread.php?t=174031

[netmask56]

Suggest you post a frame of the original - give as much detail as you can.

[idi0t]

sorry you get it all wrong.
i want to change a real video to anime video.
or want to increase the resolution.
if any of two possible please lead me.
i have attach a frame image of video as you asked for that.
[Attachment 42163 - Click to enlarge]

[idi0t]

at least tell me how to increase resolution if animation is not possible.

[jagabo]

i want to change a real video to anime video.

Look into filters like MSU Cartoonizer.
http://compression.ru/video/cartoonizer/index_en.html

[andkar]

Lika jagabo just said before I finished:

[johnmeyer]

You can also export the video as a series of still images (VirtualDub can do this). Then, use the batch capability in your photo editor (I use the ancient Photoimpact) to apply a cartoon filter to all of the individual images. Then reassemble the images back into a video. This sounds a little complicated, but it opens up a huge array of effect options that you may not find with a plugin.

However, I'd start with Cartoonizer, as already recommended, and only follow my advice if you don't get the effect you want.

As for the earlier response to my statement that a 1920x1080 video is not higher resolution than 320x240, that is simply wrong on the face of it. I guess it all gets down to how you define "resolution," but for me it is simply the number of pixels the image or video contains. Note that I did not say "high definition" because, of course, if you start with 320x240 pixels then you won't add any definition simply by scaling up. Since that is what I said in my earlier post, I have no idea why anyone would take the time to disagree.

Please note: most video cameras these days can produce 1920x1080 video. However, some of them (like my Moultrie trail camera) don't produce very sharp video. One could argue that they are not producing true HD video but, since the camera doesn't offer a setting lower than 1920x1080, and 1920x1080 is how the video is delivered, what "resolution" does it produce? I could say that it doesn't appear to produce true HD video, and I could take a video of a resolution chart and prove my claim. But the fact remains that I can only get one "flavor" of video out of this camera and that is 1920x1080.

I also note that the same person pointed to "super resolution." I suggest you search this forum (or search doom9.org) for that term and see how many (dozens and dozens) of threads have been devoted to this snake oil. You will find the famous lighthouse picture and also lots of links to various academic papers, many of them presented at Sigraph and other prestigious conferences. It is a really cool concept, and the math is intriguing.

Unfortunately, I have yet to see one instance where the result had any perceivable, actual, provable detail that was not in the original unscaled image.

Don't waste your time.

P.S. Words matter, and this exchange about "resolution" reminds me of the words we used forty years ago when I worked at the HP's microwave test and measurement division. There were multiple words used to specify the capability of measuring equipment including: accuracy, precision, and resolution. At first, they all sounded like they were describing the same thing, but in fact, each one described a different aspect of the product's ability to make a certain measurement, and they were all, each in their own way, important.

[jagabo]

I guess it all gets down to how you define "resolution,"

That's exactly the issue. It's unfortunate that the term resolution has come to mean frame size (or total number of pixels) in digital imaging. Because the number of pixels does not necessarily relate to the ability to resolve detail in that image. I prefer to use the term "frame size" rather than resolution to avoid this misunderstanding.

Yes, you can easily enlarge the frame size of a small video. Some methods are better than others. But you won't be increasing the amount of detail within the image.

[idi0t]

so i can't increasing the amount of detail within the image or video?
so there is not any way to make a video more clear and detailing?

[JVRaines]

No matter how you cook your soup, ingredients will not appear in it by magic.

[pandy]

I also note that the same person pointed to "super resolution." I suggest you search this forum (or search doom9.org) for that term and see how many (dozens and dozens) of threads have been devoted to this snake oil. You will find the famous lighthouse picture and also lots of links to various academic papers, many of them presented at Sigraph and other prestigious conferences. It is a really cool concept, and the math is intriguing.

Unfortunately, I have yet to see one instance where the result had any perceivable, actual, provable detail that was not in the original unscaled image.

Don't waste your time.

This is more tricky and stochastic sampling, time equivalent sampling are working solutions to particular class of problems - and yes, there is plenty theory behind super-resolution - i can understand your objection related to lack of easy to use tool - higher resolution source approach is cheaper and easier to implement.

[johnmeyer]

My objection has nothing to do with the ease of use. Instead, as you will find in the dozens of threads on this subject, no one has yet provide even one independently-created example which shows a super resolution result that actually improves the video beyond anything that can be done with a good upscaling algorithm. Yes, a few of the sites which describe this technology claim to show examples, but I am talking about something that you or I could do with the technology available to any one of us.

Super resolution always comes up in those "help me read this license plate" threads where people think you can extract more detail from a video of a thief's car. No one can ever read the license plate.

Therefore, I'll stick with my "snake oil" description.

[johnmeyer]

so i can't increasing the amount of detail within the image or video?
so there is not any way to make a video more clear and detailing?

That is 100% correct.

You have been fooled by Hollywood movies, as shown in this clip I've linked to many times before: Let's Enhance

Having said that, there are several techniques for fooling your eye into thinking the video has more detail. You will find this technology under the heading "sharpening." It actually reduces detail a little but, by making the transitions between dark and light (or between colors) more abrupt, or by painting those transitions with a halo (which is what the VHS tape sharpening circuits do by adding a peaking circuit) your eye will sometimes perceive the image as having more detail. However, it is an optical illusion and no detail is created. What is worse, most people use too much and the resulting video looks really bad.

[pandy]

Super resolution always comes up in those "help me read this license plate" threads where people think you can extract more detail from a video of a thief's car. No one can ever read the license plate.

Therefore, I'll stick with my "snake oil" description.

Well, once again for easy super resolution requirements are quite simple - sensor resolution can be low but framerate must be high (ideally pixels should partially overlap) - common problem with CCTV is low resolution and low framerate which prevent easy reconstruction - in real life you facing super resolution every day - sub pixel accuracy is second side of super resolution coin.

[jagabo]

so i can't increasing the amount of detail within the image or video?
so there is not any way to make a video more clear and detailing?

The best you can hope for is to increase the frame size while keeping sharp edges, and not creating aliasing artifacts and over-sharpening halos.

Here's example using a small image I created from a high definition source:



Note that this is a very clean, sharp image with no compression artifacts. So it represents a best case scenario.

Be sure to view the images below at full size (1920x1080).

Here's how a typical media player or TV would upscale the video to 1920x1080:



Notice how the sharp edges have aliasing artifacts and over-sharpening halos.

Here's the frame upscaled with one of the better upscaling techniques in AviSynth (nnedi3):



Notice that the edges are sharp and have essentially no aliasing or over-sharpening halos. There is very little detail in the orange background but the image looks pretty good. Especially compared to the hardware upscaling in the previous image.

Now here's the original high definition frame:



It has the sharpest edges and no aliasing or over-sharpening halos. But it also has lots of detail in the orange background. So we were able upscale an image and create sharp edges but we were not able to restore the detail of the original HD image.

In the real world, you are typically starting with a much worse source. A compressed video will be full of DCT ringing and blocking artifacts. Upscaling those artifacts will make them bigger and more pronounced. So you want to filter them away before you upscale. But that causes more loss of detail. So you get sharp upscaled edges with no detail between them. People look like mannequins, fuzzy sweaters look like silk shirts, small text is still unreadable, etc. Ie, crap in, crap out.

[idi0t]

thank you very much all of you
it help me a lot.