I have a video which is a little on the blurry side which i would like to make more clear and crisp if possible. I know there are many video editing programs out there with the sharpening function, but is there a specific program that does a better job than others at sharpening videos? It would be cool if i could somehow make it like blu ray quality in terms of clarity and crispness, but i know beggars cant be choosers.
+ Reply to Thread
Results 1 to 13 of 13
If it was possible to do that, nobody would buy 4K. You can't get silk purse from a sows ear. Sharpening creates all kind of weird artifacts.
I'm not too sure about the "super resolution" plugin. Has anyone used it?
I downloaded the Avisynth version and took it for a test drive. I tried upscaling several different SD videos to 1080p, and I could barely find a frame where the super resolution plugin did any better than Spline36Resize, but when I downloaded a couple of the SD sample videos from here and upscaled them, suddenly it was doing a much better job. Can an Avisynth plugin be trained to produce a better result with specific samples?
They seem to be attempting to imply the low res samples at the bottom of the page I linked to above were upscaled from the SD samples on the same page, but that isn't the case. While the Avsynth plugin did do a better job at upscaling the two SD samples I downloaded from their site than it did with my videos, the result wasn't even close to the HD samples. It couldn't be, as the SD samples look like they were created from an interlaced source that was incorrectly resized.
It looks to me like the SD images were made by downscaling the HD images. The HD images were supplied as a reference so you could perform PSNR calculations on your upscaled results. I think they downscaled using an algorithm that's especially friendly to their Super Resolution upscaling filter. It looks like a simple bilinear downscale -- similar to VirtualDub's "Bilinear (interpolation only)".
For Super Resolution to produce sharp output it needs to start with sharp input. That may be why it didn't work well with your own samples.
For the OP to use it as a sharpener he will have to downscale his images until they are sharp, then apply Super Resolution to upscale back to the original size. He should supply some samples of his source for more precise guidance. And, as has been pointed out, don't expect miracles.
I'll try downscaling my PAL video later and running it by the super resolution plugin again, although that's somewhat counter-intuitive... downscaling in order to sharpen or upscale....
I think it's so hard to reach that level of quality. You can optimize its quality via adjusting the video saturation, brightness, contrast, hue, gamma, etc. More quality improvement methods also include stabilizing shaky video, removing grain noise from clips and using filters.
@hello_hello: wouldn't it be 'better' to use super resolution to increase the size and then downscale,...?users currently on my ignore list: deadrats, Stears555
I've been scouring through the forums looking for clues to try and do some magic on an old vhs capture that has been digitized in the best manner possible by a well respected member here . I realize this is likely a fools errand, but I've also seen some pretty amazing things done with video that I simply did not think were possible, and I know there are newer "learning" algorithms that can do some things that we didn't think were possible before. I know it is not possible to turn SD content into HD content, there is simply not enough information there, but I have personally witnessed creating a video that is perceivably sharper with more clarity than it's original VHS source.
I've heard bits and peices of wisdom (mainly from jagabo..thank you!) here and there such as:
- downscale until video is sharp, then apply the SR algorithm for sharper upscaled video: As Selur said, that DOES sound counter-intuitive as you are throwing information/detail away when downscaling, correct? Just trying to understand how that ends up with a better result.
- apply de-interlace/resize, then rotate video 90 degrees and do it again. I'm sure I'm botching this one...but those in the know will correct me.
Of course the best thing I could have done is go back in time 18 years and say "nah man use a SVHS tape instead of VHS"...but a stupid 22 year old wasn't thinking much about the future
I'm almost embarrassed to ask this, but it's been a few years since I've been on here and thought I'd shake the bushes and see what falls out...
Thanks in advance for anyone who may take the time to school me here...
so a good rule of thumb would be to reduce it (probably to 50% from 720x480 to 360x240) so that is is closer to it's native "resolution" (I know that's not the right term :P ).
Perhaps applying some light sharpening filters or just letting it ride as-is using a good resize tool.
I'm using StaxRip because I'd just rather not go re-grab all the latest avisynth plugins with the endless folders, dependancies, etc.
I'm currently de-interlacing using QTGMC medium setting and assuming I didn't botch the field order setting I should end up with a smooth 60fps 720x480 video.
I will then downscale 50% using the spline64resize plugin.
This program doesn't have the huffyuv codec but has one called UTVideo that I'm told is also a lossless algorithm .
Then I'll take it from there to see what kind of damage I can do to the video
With real world video it can look very artificial because you end up with sharp edges but no real detail between them. It works much better with animated material because it starts out with sharp edges and no detail between them.
I've had great results sharpening video with HDConverttoX(portable but you will have to install a few easy things to get it working) on blurry sources. An old program, but has an incredible amount of avisynth plugins set up for you. Use the Film sharp setting in the advanced options section, if not the desired effect use the other setting Light sharp/soft sharp which will also denoise as well.