Closed nlolnlolnlolnlo closed 2 months ago
this is a problem that i noticed in ntscqt as well, is there a way to tweak the ntsc-rs sharpening to more closely match a real vcr's sharpening? it seems like it extends out a few more pixels than it should
The sharpening "size" is tied to the VHS tape speed setting--what do you currently have it at? If it's set to SP and still excessively wide, I can try to add finer control.
it is set to standard play, here's the image source & preset
ntsc-rs_vhs-sp_v6_7_smooth.json
a more noticeable difference between the real vhs images and the ntsc-rs images is how big the grayscale gradient thingy is
prob because of the bandwidth scale maybe
If you're referring to the subtle rainbow fringe, I think that's the "Input luma filter" setting's doing. It reduces crosstalk between chroma and luma when enabled.
no, im referring to the grayscale gradient thingy on the left side of the photo
it looks a little big
no, im referring to the grayscale gradient thingy on the left side of the photo
Color bleeding. It happens when the overscan of the image is casted onto the VHS signal, and that gets smeared. I'm not 100% sure on how it works, but that's from my experience
i'm not sure how helpful it is but i did some blend mode stuff and was able to approximate the vcr's maximum sharpening as an overlay
unsharpened image and extracted sharpening
seems like theres more horizitional sharpness going on than vertical sharpness. what if the sharpness option was split in horizitional and vertical?
vertical sharpening is not something that would work with interlacing
then dont split the option in half but instead make the vertical sharpening less than the horizitional sharpening
while the sharpness option still being one thing
AFAIK vertical sharpening isn't a thing in VHS because all the blurring happens horizontally, that is, per scanline. I will thus not be adding vertical sharpening, and I don't think it's what @nlolnlolnlolnlo was asking for in the first place.
Quick update on this: I think the current bandpass filter (three cascaded constant-k filters, inherited all the way back from composite-video-simulator) is to blame. I added a slider to control the cutoff for the sharpening highpass filter, but it does almost nothing.
At some point, I'll try to add the option to use a Butterworth filter or similar instead. Hopefully that'll also increase accuracy for everything else that uses a bandpass filter.
I've implemented the option to use a Butterworth filter as the low-pass filter throughout, and also made the VHS sharpening filter's frequency adjustable (since it actually does something with the Butterworth filter). I've managed to get this result: butter.json
Let me know if this suits your needs.
i completely forgot to reply but that's perfect. closing the issue now