HeadBangeR77
18th January 2007, 02:02
First of all: welcome! ;)
I've been trying hard to discover everything on my own, making lots of sample clips, comparing them and so on, but I'm still too unexperienced to understand some things. Some of you have been using those scripts (thank you Didee, you are truly a gentleman and a scholar :D) for years already, I - for a few weeks. I don't expect any all-in-one solutions, neither do I ask what's best (Read the Rules!!! :p). To the matter ...
1) Does LS(F) resize, while supersampling, the input source resolution as it is or stretches the source first to the display resolution, when dealing with anomorph MPEG-2 material? I mean: does the supersampling is done on square pixels or anomorph ones? I think the latter is right, but just want to be 100% sure (that's all of course under the condition we haven't resized before) ...
2) The higher the SS, the less artifacts but also less sharpening, as far as I can recall Didee's own words. So, is it safe to set a higher value for strength and for SS also, in order to achieve the same or even stronger sharpening effect as with lower SS, but get less artifacts (altough I can't see many ;))?
I mean sth like:
LimitedSharpenFaster(strength=80) vs. LimitedSharpenFaster(ss_x=2.0,ss_y=2.0,strenght=120)
3) Input & output resolutions. After having cropped black bars from 720x480 (yes, I've got NTSC source) to 712x360 I could either encode (just an example! I know there are more possibilities): (a) downsizing (what I usually do) a bit with LanczosResize(720x304), (b) anomorph (would have to add black bars or fill margins or sth similar to get mod16 resolution) or (c) at full display resolution, eventually upsizing a bit - that's what I'm doing atm, 1024x432.
Now to my third question, 'cause I'm not sure:
If I left LSF at default (SS x 1.5 etc.) the sharpening effect will change, won't it?
a) 712x360 => LSF SS x 1.5 (1080x540) => LSF spits out 720x304. X axis will be supersampled a bit less (assuming that LSF doesn't stretch, hence the first question above), and Y axis will be supersampled more. Am I right? Wouldn't it be better then to resize beforehand to achieve an even sharpening? (even if it would mean doing the resizing twice)
b) LSF will supersample as default, x 1.5, 'cause the resolution doesn't change - black bars are added afterwards.
c) 712x360 => LSF SS x 1.5 (1080x540) => 1024x432. So I assume the sharpening effect will be much stronger (and the possiblity of artifacts also)? I've decided to use SS x 2.0 in such a case, and that's what I'm doing now (samples look all right, so I guess I'm right ;)). Do you think it's ok?
Time for a break before dealing with PP after sharpening (including Soothe) - I will carry on later ;)
PS. Pretty please, don't start another ITU vs. generic resizing crusade! I've made dozens of samples and chosen sth in between those two, and the result satisfies my personal sense of proportions :) Don't make a mess like e.g. in that thread (http://forum.doom9.org/showthread.php?t=85220)! :p
I've been trying hard to discover everything on my own, making lots of sample clips, comparing them and so on, but I'm still too unexperienced to understand some things. Some of you have been using those scripts (thank you Didee, you are truly a gentleman and a scholar :D) for years already, I - for a few weeks. I don't expect any all-in-one solutions, neither do I ask what's best (Read the Rules!!! :p). To the matter ...
1) Does LS(F) resize, while supersampling, the input source resolution as it is or stretches the source first to the display resolution, when dealing with anomorph MPEG-2 material? I mean: does the supersampling is done on square pixels or anomorph ones? I think the latter is right, but just want to be 100% sure (that's all of course under the condition we haven't resized before) ...
2) The higher the SS, the less artifacts but also less sharpening, as far as I can recall Didee's own words. So, is it safe to set a higher value for strength and for SS also, in order to achieve the same or even stronger sharpening effect as with lower SS, but get less artifacts (altough I can't see many ;))?
I mean sth like:
LimitedSharpenFaster(strength=80) vs. LimitedSharpenFaster(ss_x=2.0,ss_y=2.0,strenght=120)
3) Input & output resolutions. After having cropped black bars from 720x480 (yes, I've got NTSC source) to 712x360 I could either encode (just an example! I know there are more possibilities): (a) downsizing (what I usually do) a bit with LanczosResize(720x304), (b) anomorph (would have to add black bars or fill margins or sth similar to get mod16 resolution) or (c) at full display resolution, eventually upsizing a bit - that's what I'm doing atm, 1024x432.
Now to my third question, 'cause I'm not sure:
If I left LSF at default (SS x 1.5 etc.) the sharpening effect will change, won't it?
a) 712x360 => LSF SS x 1.5 (1080x540) => LSF spits out 720x304. X axis will be supersampled a bit less (assuming that LSF doesn't stretch, hence the first question above), and Y axis will be supersampled more. Am I right? Wouldn't it be better then to resize beforehand to achieve an even sharpening? (even if it would mean doing the resizing twice)
b) LSF will supersample as default, x 1.5, 'cause the resolution doesn't change - black bars are added afterwards.
c) 712x360 => LSF SS x 1.5 (1080x540) => 1024x432. So I assume the sharpening effect will be much stronger (and the possiblity of artifacts also)? I've decided to use SS x 2.0 in such a case, and that's what I'm doing now (samples look all right, so I guess I'm right ;)). Do you think it's ok?
Time for a break before dealing with PP after sharpening (including Soothe) - I will carry on later ;)
PS. Pretty please, don't start another ITU vs. generic resizing crusade! I've made dozens of samples and chosen sth in between those two, and the result satisfies my personal sense of proportions :) Don't make a mess like e.g. in that thread (http://forum.doom9.org/showthread.php?t=85220)! :p