Abyssal
22nd February 2012, 21:43
Sorry for the very creative and informative title. :P
Anyways, I'm new to the forums (as you probably can see on my profile) but I'm also new to AviSynth.
So therefor I've decided to make a thread hoping to get some help with a few questions I have related to AviSynth usage.
I've been encoding a lot of anime lately and I've started to dig myself into the wondrous world of AviSynth and there are some terms I need help with.
This first one is something that generally shows up on darker areas of the encode, no - not banding but actually flat dark colors. Like a completely blue/black wall with no particularly varying color schemes.
It is some kind of gridding effect that can sometimes get a bit blocky as well.
I've cut out the area of the effect from screenshots;
Here is the raw BDMV (Sorry if the cutouts are slightly off by a few pixels but I think you'll get the idea);
http://tpspic.me/i/ThpRW.png - Yes, I know the shirt is not present and this is a different frame but it shows the wall (in reality a locker but whatever :P)
Here it is, encoded but unfiltered:
http://tpspic.me/i/K0yp.png
Here it is with some debanding and sharpening:
http://tpspic.me/i/a9Xs.png
As you can see, it's not present in the source so it's only in the encode but as you can see, debanding makes it slightly more visible and introduces some blocking to it - I did some sharpening just to make it easier to see.
So my first question - Is there any term for this so I know what to call it in the future(?) and secondly, is there any way to repair this with any particular filter?
-----
Also a sub question to this issue related to banding, is there any good explanation anywhere on how GradFun2db() and GradFun2DBmod() works? I mean the syntax - How do I properly control the threshold, radius etc. what exactly do everything do? I have seen some explanations but they are somewhat complicated and not easily understandable for someone who is completely new to AviSynth.
-----
...and my actual second question is related to something different I have come across when encoding anime and that is some kind of noise that appear around lines when encoding.
It is like dotcrawl but not as extreme - it's more subtle and seem to appear if you have "insufficient" bitrate or high crf rate so if I increase the bitrate/lower the crf rate the amount of it tends to decrease and sometimes even disappear.
But what I'm wondering if there is any filters out there that might help me reduce/remove this without increasing the quality and filesize? I do have noticed that some sharpeners tends to make this less visible - LSFmod as an example.
-----
Another thing I've noticed which becomes visible with low bitrate/high crf rate and that is lines which becomes jagged in the encode.
Here is an example of this:
Raw BDMV:
http://tpspic.me/i/xZBvh.png
Encode:
http://tpspic.me/i/CAj3V.png
This is a zoomed in cutout of a screenshot of someones hair but it can sometimes be visible on clothes and pretty much any other lines as well but I tend to see it more often on people's hair since they tend to move their head quite a bit and it is during motion it is the most visible.
But as you can see, on the lines in the encode you can see that they are somewhat jagged - I'm not sure if this can be called aliasing but whatever it is, what is usually the best way to handle this?
Regards,
Abyssal
Anyways, I'm new to the forums (as you probably can see on my profile) but I'm also new to AviSynth.
So therefor I've decided to make a thread hoping to get some help with a few questions I have related to AviSynth usage.
I've been encoding a lot of anime lately and I've started to dig myself into the wondrous world of AviSynth and there are some terms I need help with.
This first one is something that generally shows up on darker areas of the encode, no - not banding but actually flat dark colors. Like a completely blue/black wall with no particularly varying color schemes.
It is some kind of gridding effect that can sometimes get a bit blocky as well.
I've cut out the area of the effect from screenshots;
Here is the raw BDMV (Sorry if the cutouts are slightly off by a few pixels but I think you'll get the idea);
http://tpspic.me/i/ThpRW.png - Yes, I know the shirt is not present and this is a different frame but it shows the wall (in reality a locker but whatever :P)
Here it is, encoded but unfiltered:
http://tpspic.me/i/K0yp.png
Here it is with some debanding and sharpening:
http://tpspic.me/i/a9Xs.png
As you can see, it's not present in the source so it's only in the encode but as you can see, debanding makes it slightly more visible and introduces some blocking to it - I did some sharpening just to make it easier to see.
So my first question - Is there any term for this so I know what to call it in the future(?) and secondly, is there any way to repair this with any particular filter?
-----
Also a sub question to this issue related to banding, is there any good explanation anywhere on how GradFun2db() and GradFun2DBmod() works? I mean the syntax - How do I properly control the threshold, radius etc. what exactly do everything do? I have seen some explanations but they are somewhat complicated and not easily understandable for someone who is completely new to AviSynth.
-----
...and my actual second question is related to something different I have come across when encoding anime and that is some kind of noise that appear around lines when encoding.
It is like dotcrawl but not as extreme - it's more subtle and seem to appear if you have "insufficient" bitrate or high crf rate so if I increase the bitrate/lower the crf rate the amount of it tends to decrease and sometimes even disappear.
But what I'm wondering if there is any filters out there that might help me reduce/remove this without increasing the quality and filesize? I do have noticed that some sharpeners tends to make this less visible - LSFmod as an example.
-----
Another thing I've noticed which becomes visible with low bitrate/high crf rate and that is lines which becomes jagged in the encode.
Here is an example of this:
Raw BDMV:
http://tpspic.me/i/xZBvh.png
Encode:
http://tpspic.me/i/CAj3V.png
This is a zoomed in cutout of a screenshot of someones hair but it can sometimes be visible on clothes and pretty much any other lines as well but I tend to see it more often on people's hair since they tend to move their head quite a bit and it is during motion it is the most visible.
But as you can see, on the lines in the encode you can see that they are somewhat jagged - I'm not sure if this can be called aliasing but whatever it is, what is usually the best way to handle this?
Regards,
Abyssal